SlideShare a Scribd company logo

Responsible AI in Industry (ICML 2021 Tutorial)

Download as PPTX, PDF
••
K
Krishnaram Kenthapadi

The document outlines practical challenges and lessons learned in implementing responsible AI in various industries, focusing on ethical issues, algorithmic bias, and the importance of fairness, transparency, and explainability in AI systems. It discusses different tools and methodologies for fairness-aware machine learning, highlighting case studies from companies like Amazon, Google, Microsoft, and Philips on ensuring fairness in AI applications. Key takeaways include the necessity of testing for biases, improving models through better data handling, and establishing trust in AI by meeting regulatory compliance.

Internet
1 of 198
Downloaded 55 times
1
2
Most read
3
4
5
6
Most read
7
8
9
10
11
12
13
14
15
16
17
18
19
Most read
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
Responsible AI in Industry: Practical Challenges and Lessons Learned TutoriaI 2021 Krishnaram Kenthapadi (Amazon AWS AI), Ben Packer (Google AI), Mehrnoosh Sameki (Microsoft Azure), Nashlie Sephus (Amazon AWS AI) https://sites.google.com/view/ResponsibleAITutorial
Massachusetts Group Insurance Commission (1997): Anonymized medical history of state employees William Weld vs Latanya Sweeney Latanya Sweeney (MIT grad student): $20 – Cambridge voter roll born July 31, 1945 resident of 02138
64% Uniquely identifiable with ZIP + birth date + gender (in the US population) Golle, “Revisiting the Uniqueness of Simple Demographics in the US Population”, WPES 2006
A History of Privacy Failures … Credit: Kobbi Nissim,Or Sheffet
Responsible AI in Industry (ICML 2021 Tutorial)
Responsible AI in Industry (ICML 2021 Tutorial)
• Ethical challenges posed by AI systems • Inherent biases present in society • Reflected in training data • AI/ML models prone to amplifying such biases Algorithmic Bias
Laws against Discrimination Immigration Reform and Control Act Citizenship Rehabilitation Act of 1973; Americans with Disabilities Act of 1990 Disability status Civil Rights Act of 1964 Race Age Discrimination in Employment Act of 1967 Age Equal Pay Act of 1963; Civil Rights Act of 1964 Sex And more...
Fairness Privacy Transparency Explainability
Motivation & Business Opportunities Regulatory. We need to understand why the ML model made a given prediction and also whether the prediction it made was free from bias, both in training and at inference. Business. Providing explanations to internal teams (loan officers, customer service rep, compliance teams) and end users/customers Data Science. Improving models through better feature engineering and training data generation, understanding failure modes of the model, debugging model predictions, etc.
© 2020, Amazon Web Services, Inc. or its Affiliates. Amazon SageMaker VISION SPEECH TEXT SEARCH CHATBOTS PERSONALIZATION FORECASTING FRAUD CONTACT CENTERS Deep Learning AMIs & Containers GPUs & CPUs Elastic Inference Trainium Inferentia FPGA AI SERVICES ML SERVICES FRAMEWORKS & INFRASTRUCTURE DeepGraphLibrary Amazon Rekognition Amazon Polly Amazon Transcribe +Medical Amazon Lex Amazon Personalize Amazon Forecast Amazon Comprehend +Medical Amazon Textract Amazon Kendra Amazon CodeGuru Amazon Fraud Detector Amazon Translate INDUSTRIAL AI CODE AND DEVOPS NEW Amazon DevOps Guru Voice ID For AmazonConnect Contact Lens NEW Amazon Monitron NEW AWS Panorama + Appliance NEW Amazon Lookout for Vision NEW Amazon Lookout for Equipment Scaling Fairness, Explainability & Privacy across the AWS ML Stack NEW Amazon HealthLake HEALTHCARE AI NEW Amazon Lookout for Metrics ANOMALY DETECTION Amazon Transcribe for Medical Amazon Comprehend for Medical Label data NEW Aggregate & prepare data NEW Store & share features Auto ML Spark/R NEW Detect bias Visualize in notebooks Pick algorithm Train models Tune parameters NEW Debug & profile Deploy in production Manage & monitor NEW CI/CD Human review review NEW: Model management for edge devices NEW: SageMaker JumpStart SAGEMAKER STUDIO IDE
LinkedIn operates the largest professional network on the Internet Tell your story 740M members 55M+ companies are represented on LinkedIn 90K schools listed (high school & college) 36K skills listed 14M+ open jobs on LinkedIn Jobs 280B Feed updates
Tutorial Outline • Fairness-aware ML: An overview • Explainable AI: An overview • Privacy-preserving ML: An overview • Responsible AI tools • Case studies • Key takeaways
Fairness-aware ML: An Overview Nashlie Sephus, PhD Applied Science Manager, AWS AI
• ML Fairness Considerations • ML and Humans • What is fairness/inclusion for ML? • Where May Biases Occur? • Testing Techniques with Face Experiments • Takeaways Outline 15
Product Introspection (1): Make Your Key Choices Explicit [Mitchell et al., 2018] Goals Decision Prediction Profit from loans Whether to lend Loan will be repaid Justice, Public safety Whether to detain Crime committed if not detained • Goals are ideally measurable • What are your non-goals? • Which decisions are you not considering? • What is the relationship between Prediction and Decision?
Product Introspection (2): Identify Potential Harms • What are the potential harms? • Applicants who would have repaid are not given loans • Convicts who would not commit a crime are locked up. • Are there also longer term harms? • Applicants are given loans, then go on to default, harming their credit score • Are some harms especially bad?
Seek out Diverse Perspectives • Fairness Experts • User Researchers • Privacy Experts • Legal • Social Science Backgrounds • Diverse Identities • Gender • Sexual Orientation • Race • Nationality • Religion
AI/ML and Humans 19
! 16
Responsible AI in Industry (ICML 2021 Tutorial)
22
Error-free (no system is perfect) 100% confident Intended to replace human judgement What ML Is Not 23
Responsible AI in Industry (ICML 2021 Tutorial)
Responsible AI in Industry (ICML 2021 Tutorial)
Fairness Techniques in Faces 26
Detect presence of a face in an image or a video. Face Detection 27
A system to determine the gender, age, emotion, presence of facial hair, etc. from a detected face. Face Analysis
A system to determine a detected faces identity by matching it against a database of faces and their associated identities. Face Recognition 29
Estimation of the confidence or certainty of any prediction Expressed in the form of a probability or confidence score Confidence Score 30
Face Recognition: Common Causes of Errors ILLUMINATION VARIANCE POSE / VIEWPOINT AGING EXPRESSION / STYLE OCCLUSION Lighting, camera controls like exposure, shadows, highlights Face pose, camera angles Natural aging, artificial makeup Face expression like laughing, facial hair such as a beard, hair style Part of the face hidden as in group pictures 31
Where Can Biases Exist? 32
AFRICAN SCANDINAVIAN A B C The PPB dataset [Buolamwini 2018] 6.3% 20.8% GenderShades.Org 33 [Buolamwini & Gebru 2018]
Darker Skinned Female 34
Racial Comparisons of Datasets [FairFace] 35
Black Male
Latino Hispanic Male 37
Southeast Asian Female 38
[Nvidia] 39
Launch with Confidence: Testing for Bias • How will you know if users are being harmed? • How will you know if harms are unfairly distributed? • Detailed testing practices are often not covered in academic papers • Discussing testing requirements is a useful focal point for cross-functional teams
Reproducibility - Notebook Experiments 41
PPB2 Data Analytics 42
43
Gender Classification – PPB2 44
45
46
47
48
Short Hair 49
Gender Classification w.r.t. Hair Lengths – PPB2 50
FairFace Dataset Analytics 51
Gender classification – FairFace 52
Model Predictions Evaluate for Inclusion - Confusion Matrix
Model Predictions Positive Negative Evaluate for Inclusion - Confusion Matrix
Model Predictions Positive Negative ● Exists ● Predicted True Positives ● Doesn’t exist ● Not predicted True Negatives Evaluate for Inclusion - Confusion Matrix
Model Predictions Positive Negative ● Exists ● Predicted True Positives ● Exists ● Not predicted False Negatives ● Doesn’t exist ● Predicted False Positives ● Doesn’t exist ● Not predicted True Negatives Evaluate for Inclusion - Confusion Matrix
Efficient Testing for Bias • Development teams are under multiple constraints • Time • Money • Human resources • Access to data • How can we efficiently test for bias? • Prioritization • Strategic testing
Choose your evaluation metrics in light of acceptable tradeoffs between False Positives and False Negatives
Takeaways • Testing for blindspots amongst intersectionality is key. • Taking into account confidence scores/thresholds and error bars when measuring for biases is necessary. • Representation matters. • Transparency, reproducibility, and education can promote change. • Confidence in your product's fairness requires fairness testing • Fairness testing has a role throughout the product iteration lifecycle • Contextual concerns should be used to prioritize fairness testing 59
AI Fairness and Transparency Tools Mehrnoosh Sameki Microsoft Azure
Microsoft
Machine Learning Transparency and Fairness
Interpretability
Understand and debug your model Glassbox Models: Model Types: Linear Models, Decision Trees, Decision Rules, Explainable Boosting Machines Blackbox Models: Model Formats: Python models using scikit predict convention, Scikit, Tensorflow, Pytorch, Keras, Explainers: SHAP, LIME, Global Surrogate, Feature Permutation DiCE Diverse Counterfactual Explanations Interpret Glassbox and blackbox interpretability methods for tabular data Interpret-text Interpretability methods for text data Interpret-community Community-driven interpretability techniques for tabular data https://github.com/interpretml Azureml-interpret AzureML SDK wrapper for Interpret and Interpret-community
Interpretability Approaches
Models designed to be interpretable. Lossless explainability. Fever? Internal Bleeding? Stay Home Stay Home Go to Hospital Decision Trees Rule Lists Linear Models ….
Explain any ML system. Approximate explainability. Model Explanation Perturb Inputs Analyze SHAP LIME Partial Dependence Sensitivity Analysis
Fairness Useful links: • AI Show • Tutorial Video • Customer Highlight
There are many ways that an AI system can behave unfairly. Fairness in AI Avoiding negative outcomes of AI systems for different groups of people A model for screening loan or job application might be much better at picking good candidates among white men than among other groups. A voice recognition system might fail to work as well for women as it does for men.
https://github.com/fairlearn/fairlearn Assessing unfairness in your model Fairness Assessment: Usecommonfairness metricsandaninteractive dashboardto assess which groups of peoplemay benegatively impacted. Model Formats: Python models using scikit predict convention, Scikit, Tensorflow, Pytorch, Keras Metrics: 15+ Common group fairness metrics Model Types: Classification, Regression Unfairness Mitigation: Use state-of-the-art algorithmsto mitigate unfairness in yourclassificationandregressionmodels.
Fairness Assessment Input Selections Sensitive attribute Performance metric Assessment Results Disparity in performance Disparity in predictions Mitigation Algorithms Post-processing algorithm Reductions Algorithm
Situation: Solution: Philips Healthcare used Fairlearn to check whether our ICU models perform similarly for patients with different ethnicities and gender identities, etc. Microsoft CSE (Led by Tempest Van Schaik) collaborated with Philips to build a solution using Azure DevOps pipelines, Azure Databricks and Mlflow. Built a pipeline to make fairness monitoring routine, checking the fairness of predictions for patients of different genders, ethnicities, and medical conditions, using Fairlearn metrics. Fairness analysis helped show that Philips’ predictive model performs better than industry standard ICU models Standard model predictions for a patient differ depending on how the ICU documented their test results. Customer: Philips Industry: Healthcare Size: 80,000+ employees Country: Netherlands Products and services: Microsoft Azure DevOps Microsoft Azure Databricks MLFlow Putting fairness monitoring in production with ICU models Philips Healthcare Informatics Philips Healthcare Informatics helps ICUs benchmark their performance (e.g. mortality rate). They create quarterly benchmark reports that compare actual performance vs performance predicted by ML models. They have models trained on the largest ICU dataset in USA: 400+ ICUs, 6M+ patient stays, billions of vital signs & lab tests. Deploying ICU models responsibly Philips needed a scalable, reliable, repeatable and responsible way to bring ML models into production.
Situation: Solution: Impact: “Azure Machine Learning and its Fairlearn capabilities offer advanced fairness and explainability that have helped us deploy trustworthy AI solutions for our customers, while enabling stakeholder confidence and regulatory compliance.” —Alex Mohelsky, Partner and Advisory Data, Analytic, and AI Leader, EY Canada Customer: EY Industry: Partner Professional Services Size: 10,000+ employees Country: United Kingdom Products and services: Microsoft Azure Microsoft Azure Machine Learning Read full story here Organizations won’t fully embrace AI until they trust it. EY wanted to help its customers embrace AI to help them better understand their customers, identify fraud and security breaches sooner, and make loan decisions faster and more efficiently. The company developed its EY Trusted AI Platform, which uses Microsoft Azure Machine Learning capabilities to assess and mitigate unfairness in machine learning models. Running on Azure, the platform uses Fairlearn and InterpretML, open-source capabilities in Azure Machine Learning. When EY tested Fairlearn with real mortgage data, it reduced the accuracy disparity between men and women approved or denied loans from 7 percent to less than 0.5 percent. Through this platform, EY helps customers and regulators alike gain confidence in AI and machine learning.
TECH SOCIETY Reception & Adoption Educational materials are key for adoption of fairness toolkits  manuals, case studies, white papers [Lee & Singh, 2020]
• Interpretability at training time • Combination of glass-box models and black-box explainers • Auto reason code generation for local predictions • Ability to cross reference to other techniques to ensure stability and consistency in results H2O
Responsible AI in Industry (ICML 2021 Tutorial)
IBM Open Scale Goal: to provide AI operations team with a toolkit that allows for: • Monitoring and re-evaluating machine learning models after deployment
IBM Open Scale Goal: to provide AI operations team with a toolkit that allows for: • Monitoring and re-evaluating machine learning models after deployment • ACCURACY • FAIRNESS • PERFORMANCE
IBM Open Scale
IBM Open Scale
IBM Open Scale Fairness Accuracy Performance
IBM Open Scale
AI Fairness 360 Datasets Toolbox Fairness metrics (30+) Bias mitigation algorithms (9+) Guidance Industry-specific tutorials
AI Fairness 360 Datasets Toolbox Fairness metrics (30+) Bias mitigation algorithms (9+) Guidance Industry-specific tutorials Pre-processing algorithm: a bias mitigation algorithm that is applied to training dat In-processing algorithm: a bias mitigation algorithm that is applied to a model during its training Post-processing algorithm: a bias mitigation algorithm that is applied to predicted labels
What If Tool Goal: Code-free probing of machine learning models • Feature perturbations (what if scenarios) • Counterfactual example analysis • [Classification] Explore the effects of different classification thresholds, taking into account constraints such as different numerical fairness metrics.
What If Tool
Responsible AI in Industry (ICML 2021 Tutorial)
Responsible AI in Industry (ICML 2021 Tutorial)
Datasheets for Datasets [Gebru et al., 2018] • Better data-related documentation • Datasheets for datasets: every dataset, model, or pre-trained API should be accompanied by a data sheet that documents its • Creation • Intended uses • Limitations • Maintenance • Legal and ethical considerations • Etc.
Model Cards for Model Reporting[Mitchell et al., 2018]
Fact Sheets [Arnold et al., 2019] • Is distinguished from “model cards” and “datasheets” in that the focus is on the final AI service: • What is the intended use of the service output? • What algorithms or techniques does this service implement? • Which datasets was the service tested on? (Provide links to datasets that were used for testing, along with corresponding datasheets.) • Describe the testing methodology. • Describe the test results. • Etc.
Responsible AI Case Studies at LinkedIn Krishnaram Kenthapadi Amazon AWS AI
Fairness in AI @ LinkedIn Fairness-aware Talent Search Ranking* *Work done while at LinkedIn
Guiding Principle: “Diversity by Design”
Insights to Identify Diverse Talent Pools Representative Talent Search Results Diversity Learning Curriculum “Diversity by Design” in LinkedIn’s Talent Solutions
Plan for Diversity
Representative Ranking for Talent Search S. C. Geyik, S. Ambler, K. Kenthapadi, Fairness- Aware Ranking in Search & Recommendation Systems with Application to LinkedIn Talent Search, KDD’19. [Microsoft’s AI/ML conference (MLADS’18). Distinguished Contribution Award] Building Representative Talent Search at LinkedIn (LinkedIn engineering blog)
Intuition for Measuring and Achieving Representativeness Ideal: Top ranked results should follow a desired distribution on gender/age/… E.g., same distribution as the underlying talent pool Inspired by “Equal Opportunity” definition [Hardt et al, NeurIPS’16] Defined measures (skew, divergence) based on this intuition
Measuring (Lack of) Representativeness Skew@k (Logarithmic) ratio of the proportion of candidates having a given attribute value among the top k ranked results to the corresponding desired proportion Variants: MinSkew: Minimum over all attribute values MaxSkew: Maximum over all attribute values Normalized Discounted Cumulative Skew Normalized Discounted Cumulative KL-divergence
Fairness-aware Reranking Algorithm (Simplified) Partition the set of potential candidates into different buckets for each attribute value Rank the candidates in each bucket according to the scores assigned by the machine-learned model Merge the ranked lists, balancing the representation requirements and the selection of highest scored candidates Representation requirement: Desired distribution on gender/age/… Algorithmic variants based on how we choose the next attribute
Architecture
Validating Our Approach Gender Representativeness Over 95% of all searches are representative compared to the qualified population of the search Business Metrics A/B test over LinkedIn Recruiter users for two weeks No significant change in business metrics (e.g., # InMails sent or accepted) Ramped to 100% of LinkedIn Recruiter users worldwide
Lessons learned • Post-processing approach desirable • Model agnostic • Scalable across different model choices for our application • Acts as a “fail-safe” • Robust to application-specific business logic • Easier to incorporate as part of existing systems • Build a stand-alone service or component for post-processing • No significant modifications to the existing components • Complementary to efforts to reduce bias from training data & during model training • Collaboration/consensus across key
Lessons learned • Post-processing approach desirable • Model agnostic • Scalable across different model choices for our application • Acts as a “fail-safe” • Robust to application-specific business logic • Easier to incorporate as part of existing systems • Build a stand-alone service or component for post-processing • No significant modifications to the existing components • Complementary to efforts to reduce bias from training data & during model training • Collaboration/consensus across key
Evaluating Fairness Using Permutations Tests [DiCiccio, Vasudevan, Basu, Kenthapadi, Agarwal, KDD’20] • Is the measured discrepancy across different groups statistically significant? • Use statistical hypothesis tests! • Can we perform hypothesis tests in a metric-agnostic manner? • Non-parametric tests can help! • Permutation testing framework
Brief Review of Permutation Tests Observe data from two populations: and Are the populations the same? A reasonable test statistic might be
Brief Review of Permutation Tests (Continued) A p-value is the chance of observing a test statistic at least as “extreme” as the value we actually observed Permutation test approach: ●Randomly shuffle the population designations of the observations ●Recompute the test statistic T ●Repeat many times Permutation p-value: the proportion of permuted datasets resulting in a larger test statistic than the original value This test is exact!
A Fairness Example Consider testing whether the true positive rate of a classifier is equal between two groups Test Statistic: difference in proportion of negative labeled observations that are classified as positive between the two groups Permutation test: Randomly reshuffle group labels, recompute test statistic
Permutations Tests for Evaluating Fairness in ML Models • Issues with classical permutation test • Want to check: just equality of the fairness metric (e.g., false positive rate) across groups, and not if the two groups have identical distribution • Exact for the strong null hypothesis … • … but may not be valid (even asymptotically) for the weak null hypothesis • Our paper: A fix for this issue • Choose a pivotal statistic (asymptotically distribution-free; does not depend on the observed data’s distribution) • E.g., Studentize the test statistic
Permutations Tests for Evaluating Fairness in ML Models • Issues with classical permutation test • Want to check: just equality of the fairness metric (e.g., false positive rate) across groups, and not if the two groups have identical distribution • Exact for the strong null hypothesis … • … but may not be valid (even asymptotically) for the weak null hypothesis • Our paper: A fix for this issue • Choose a pivotal statistic (asymptotically distribution-free; does not depend on the observed data’s distribution) • E.g., Studentize the test statistic
Engineering for Fairness in AI Lifecycle S.Vasudevan, K. Kenthapadi, LiFT: A Scalable Framework for Measuring Fairness in ML Applications, CIKM’20 https://github.com/linkedin/LiFT
LiFT System Architecture [Vasudevan & Kenthapadi, CIKM’20] •Flexibility of Use (Platform agnostic) •Ad-hoc exploratory analyses •Deployment in offline workflows •Integration with ML Frameworks •Scalability •Diverse fairness metrics •Conventional fairness metrics •Benefit metrics •Statistical tests
LiFT System Architecture [Vasudevan & Kenthapadi, CIKM’20] •Flexibility of Use (Platform agnostic) •Ad-hoc exploratory analyses •Deployment in offline workflows •Integration with ML Frameworks •Scalability •Diverse fairness metrics •Conventional fairness metrics •Benefit metrics •Statistical tests
Acknowledgements LinkedIn Talent Solutions Diversity team, Hire & Careers AI team, Anti-abuse AI team, Data Science Applied Research team Special thanks to Deepak Agarwal, Parvez Ahammad, Stuart Ambler, Kinjal Basu, Jenelle Bray, Erik Buchanan, Bee-Chung Chen, Fei Chen, Patrick Cheung, Gil Cottle, Cyrus DiCiccio, Patrick Driscoll, Carlos Faham, Nadia Fawaz, Priyanka Gariba, Meg Garlinghouse, Sahin Cem Geyik, Gurwinder Gulati, Rob Hallman, Sara Harrington, Joshua Hartman, Daniel Hewlett, Nicolas Kim, Rachel Kumar, Monica Lewis, Nicole Li, Heloise Logan, Stephen Lynch, Divyakumar Menghani, Varun Mithal, Arashpreet Singh Mor, Tanvi Motwani, Preetam Nandy, Lei Ni, Nitin Panjwani, Igor Perisic, Hema Raghavan, Romer Rosales, Guillaume Saint-Jacques, Badrul Sarwar, Amir Sepehri, Arun Swami, Ram Swaminathan, Grace Tang, Ketan Thakkar, Sriram Vasudevan, Janardhanan Vembunarayanan, James Verbus, Xin Wang, Hinkmond Wong, Ya Xu, Lin Yang, Yang Yang, Chenhui Zhai, Liang Zhang, Yani Zhang
Privacy in AI @ LinkedIn PriPeARL: Framework to compute robust, privacy-preserving analytics
Analytics & Reporting Products at LinkedIn Profile View Analytics 116 Content Analytics Ad Campaign Analytics All showing demographics of members engaging with the product
Admit only a small # of predetermined query types Querying for the number of member actions, for a specified time period, together with the top demographic breakdowns Analytics & Reporting Products at LinkedIn
Admit only a small # of predetermined query types Querying for the number of member actions, for a specified time period, together with the top demographic breakdowns Analytics & Reporting Products at LinkedIn E.g., Title = “Senior Director” E.g., Clicks on a given ad
Privacy Requirements Attacker cannot infer whether a member performed an action E.g., click on an article or an ad Attacker may use auxiliary knowledge E.g., knowledge of attributes associated with the target member (say, obtained from this member’s LinkedIn profile) E.g., knowledge of all other members that performed similar action (say, by creating fake accounts)
Possible Privacy Attacks 120 Targeting: Senior directors in US, who studied at Cornell Matches ~16k LinkedIn members → over minimum targeting threshold Demographic breakdown: Company = X May match exactly one person → can determine whether the person clicks on the ad or not Require minimum reporting threshold Attacker could create fake profiles! E.g. if threshold is 10, create 9 fake profiles that all click. Rounding mechanism E.g., report incremental of 10 Still amenable to attacks E.g. using incremental counts over time to infer individuals’ actions Need rigorous techniques to preserve member privacy (not reveal exact aggregate counts)
Problem Statement Compute robust, reliable analytics in a privacy- preserving manner, while addressing the product needs.
PriPeARL: A Framework for Privacy-Preserving Analytics K. Kenthapadi, T. T. L. Tran, ACM CIKM 2018 122 Pseudo-random noise generation, inspired by differential privacy ● Entity id (e.g., ad creative/campaign/account) ● Demographic dimension ● Stat type (impressions, clicks) ● Time range ● Fixed secret seed Uniformly Random Fraction ● Cryptographic hash ● Normalize to (0,1) Random Noise Laplace Noise ● Fixed ε True Count Noisy Count To satisfy consistency requirements ● Pseudo-random noise → same query has same result over time, avoid averaging attack. ● For non-canonical queries (e.g., time ranges, aggregate multiple entities) ○ Use the hierarchy and partition into canonical queries ○ Compute noise for each canonical queries and sum up the noisy counts
PriPeARL System Architecture
Lessons Learned from Deployment (> 1 year) Semantic consistency vs. unbiased, unrounded noise Suppression of small counts Online computation and performance requirements Scaling across analytics applications Tools for ease of adoption (code/API library, hands-on how-to tutorial) help! Having a few entry points (all analytics apps built over Pinot)  wider adoption
Summary Framework to compute robust, privacy-preserving analytics Addressing challenges such as preserving member privacy, product coverage, utility, and data consistency Future Utility maximization problem given constraints on the ‘privacy loss budget’ per user E.g., noise with larger variance to impressions but less noise to clicks (or conversions) E.g., more noise to broader time range sub-queries and less noise to granular time range sub-queries Reference: K. Kenthapadi, T. Tran, PriPeARL: A Framework for Privacy- Preserving Analytics and Reporting at LinkedIn, ACM CIKM 2018.
Acknowledgements Team: AI/ML: Krishnaram Kenthapadi, Thanh T. L. Tran Ad Analytics Product & Engineering: Mark Dietz, Taylor Greason, Ian Koeppe Legal / Security: Sara Harrington, Sharon Lee, Rohit Pitke Acknowledgements Deepak Agarwal, Igor Perisic, Arun Swami
LinkedIn Salary
LinkedIn Salary (launched in Nov, 2016)
Data Privacy Challenges Minimize the risk of inferring any one individual’s compensation data Protection against data breach No single point of failure
Problem Statement How do we design LinkedIn Salary system taking into account the unique privacy and security challenges, while addressing the product requirements? K. Kenthapadi, A. Chudhary, and S. Ambler, LinkedIn Salary: A System for Secure Collection and Presentation of Structured Compensation Insights to Job Seekers, IEEE PAC 2017 (arxiv.org/abs/1705.06976)
Title Region $$ User Exp Designer SF Bay Area 100K User Exp Designer SF Bay Area 115K ... ... ... Title Region $$ User Exp Designer SF Bay Area 100K De-identification Example Title Region Company Industry Years of exp Degree FoS Skills $$ User Exp Designer SF Bay Area Google Internet 12 BS Interactive Media UX, Graphics, ... 100K Title Region Industry $$ User Exp Designer SF Bay Area Internet 100K Title Region Years of exp $$ User Exp Designer SF Bay Area 10+ 100K Title Region Company Years of exp $$ User Exp Designer SF Bay Area Google 10+ 100K #data points > threshold? Yes ⇒ Copy to Hadoop (HDFS) Note: Original submission stored as encrypted objects.
System Architecture
Acknowledgements Team: AI/ML: Krishnaram Kenthapadi, Stuart Ambler, Xi Chen, Yiqun Liu, Parul Jain, Liang Zhang, Ganesh Venkataraman, Tim Converse, Deepak Agarwal Application Engineering: Ahsan Chudhary, Alan Yang, Alex Navasardyan, Brandyn Bennett, Hrishikesh S, Jim Tao, Juan Pablo Lomeli Diaz, Patrick Schutz, Ricky Yan, Lu Zheng, Stephanie Chou, Joseph Florencio, Santosh Kumar Kancha, Anthony Duerr Product: Ryan Sandler, Keren Baruch Other teams (UED, Marketing, BizOps, Analytics, Testing, Voice of Members, Security, …): Julie Kuang, Phil Bunge, Prateek Janardhan, Fiona Li, Bharath Shetty, Sunil Mahadeshwar, Cory Scott, Tushar Dalvi, and team Acknowledgements David Freeman, Ashish Gupta, David Hardtke, Rong Rong, Ram
Responsible AI Case Studies at Amazon Krishnaram Kenthapadi Amazon AWS AI
135 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | 135 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | Detect bias in ML models and understand model predictions Amazon SageMaker Clarify
136 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | Predictive Maintenance Manufacturing, Automotive, IoT Demand Forecasting Retail, Consumer Goods, Manufacturing Fraud Detection Financial Services, Online Retail Credit Risk Prediction Financial Services, Retail Extract and Analyze Data from Documents Healthcare, Legal, Media/Ent, Education Computer Vision Healthcare, Pharma, Manufacturing Autonomous Driving Automotive, Transportation Personalized Recommendations Media & Entertainment, Retail, Education Churn Prediction Retail, Education, Software & Internet https://aws.amazon.c om/sagemaker/gettin g-started Amazon SageMaker Customer ML Use cases
137 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | Bias and Explainability: Challenges 1 Without detection, it is hard to know if bias has entered an ML model: • Imbalances may be present in the initial dataset • Bias may develop during training • Bias may develop over time after model deployment 2 Machine learning models are often complex & opaque, making explainability critical: • Regulations may require companies to be able to explain model predictions • Internal stakeholders and customers may need explanations for model behavior • Data science teams can improve models if they understand model behavior
138 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | Amazon SageMaker Clarify Detect bias in ML models and understand model predictions Detect bias during data preparation Identify imbalances in data Evaluate the degree to which various types of bias are present in your model Check your trained model for bias Understand the relative importance of each feature to your model’s behavior Explain overall model behavior Understand the relative importance of each feature for individual inferences Explain individual predictions Provide alerts and detect drift over time due to changing real-world conditions Detect drift in bias and model behavior over time Generated automated reports Produce reports on bias and explanations to support internal presentations
147 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | SageMaker Clarify Use Cases Regulatory Compliance Internal Reporting Operational Excellence Customer Service
Lessons learned • Fairness as a Process • Notions of bias & fairness are highly application dependent • Choice of the attribute(s) for which bias is to be measured & the choice of the bias metrics to be guided by social, legal, and other non-technical considerations • Collaboration/consensus across key stakeholders • Wide spectrum of customers with different levels of technical background • Managed service vs. open source packages • Monitoring of the deployed model • Fairness & explainability considerations across the ML lifecycle
150 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | Fairness and Explainability by Design in the ML Lifecycle
151 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | Additional Pointers For more information on Amazon SageMaker Clarify, please refer: • https://aws.amazon.com/sagemaker/clarify • Amazon Science / AWS Articles • https://aws.amazon.com/blogs/aws/new-amazon-sagemaker-clarify-detects- bias-and-increases-the-transparency-of-machine-learning-models • https://www.amazon.science/latest-news/how-clarify-helps-machine-learning- developers-detect-unintended-bias • Technical papers: (1) Amazon SageMaker Clarify [KDD’21] (2) Fairness Measures for Machine Learning in Finance • https://github.com/aws/amazon-sagemaker-clarify Acknowledgments: Amazon SageMaker Clarify core team, Amazon AWS AI team, and partners across Amazon
152 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | Additional Pointers For more information on Amazon SageMaker Clarify, please refer: • https://aws.amazon.com/sagemaker/clarify • Amazon Science / AWS Articles • https://aws.amazon.com/blogs/aws/new-amazon-sagemaker-clarify-detects- bias-and-increases-the-transparency-of-machine-learning-models • https://www.amazon.science/latest-news/how-clarify-helps-machine-learning- developers-detect-unintended-bias • Technical papers: (1) Amazon SageMaker Clarify [KDD’21] (2) Fairness Measures for Machine Learning in Finance • https://github.com/aws/amazon-sagemaker-clarify Acknowledgments: Amazon SageMaker Clarify core team, Amazon AWS AI team, and partners across Amazon
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Trademark Amazon SageMaker Debugger Debug and profile ML model training and get real-time insights
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Trademark Why debugging and profiling Training bugs Large compute instances Long training times Training ML models is difficult and compute intensive
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Trademark SageMaker Debugger Real-time monitoring Relevant data capture Automatic error detection SageMaker Studio integration Debug data while training is ongoing Zero code change Persistent in your S3 bucket Built-in and custom rules Early termination Alerts about rule status Save time and cost Find issues early Accelerate prototyping Detect performance bottlenecks View suggestions on resolving bottlenecks, Interactive visualizations Monitor utilization Profile by step or time duration Right size instance Improve utilization Reduce cost System resource usage Time spent by training operations
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Trademark https://www.amazon.science/publications/amazon-sagemaker-debugger-a-system-for-real-time-insights-into-machine-learning-model- training
157 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | Additional Pointers For more information on Amazon SageMaker Debugger, please refer: • https://aws.amazon.com/sagemaker/debugger • AWS Articles • https://aws.amazon.com/blogs/aws/amazon-sagemaker-debugger-debug-your- machine-learning-models • https://aws.amazon.com/blogs/machine-learning/detecting-hidden-but-non-trivial- problems-in-transfer-learning-models-using-amazon-sagemaker-debugger • Technical paper: Amazon SageMaker Debugger: A System for Real-Time Insights into Machine Learning Model Training (MLSys 2021) • https://pypi.org/project/smdebug Acknowledgments: Amazon SageMaker Debugger core team, Amazon AWS AI team, and partners across Amazon
Fairness for Opaque Models via Model Tuning (Hyperparameter Optimization) • Can we tune the hyperparameters of a model to achieve both accuracy and fairness? • Can we support both opaque models and opaque fairness constraints? • Use Bayesian optimization for HPO with fairness constraints! • Explore hyperparameter configurations where fairness constraints are satisfied V. Perrone, M. Donini, M. B. Zafar, R. Schmucker, K. Kenthapadi, C. Archambeau, Fair Bayesian Optimization, AIES 2021 (Best paper award @ ICML 2020 AutoML workshop)
Human-in-the-loop frameworks Desirable to augment ML model predictions with expert inputs Popular examples – Healthcare models Content moderation tasks Useful for improving accuracy, incorporating additional information, and auditing models. Child maltreatment hotline screening
Errors and biases in human-in-the-loop frameworks ML tasks often suffer from group-specific bias, induced due to misrepresentative data or models. Human-in-the-loop frameworks can reflect biases or inaccuracies of the human experts. Concerns include: • Racial bias in human-in-the-loop framework for recidivism risk assessment (Green, Chen – FAccT 2019) • Ethical concerns regarding audits of commercial facial processing technologies (Raji et al. – AIES 2020) • Automation bias in time critical decision support systems (Cummings – ISTC 2004) Can we design human-in-the-loop frameworks that take into account the expertise and biases of the human experts?
Model 𝑋 − non-protected attributes; 𝑌 − class label; 𝑍 − protected attribute/group membership Number of experts available = 𝑚 − 1 Input −𝑋 Classifier 𝐹: 𝑋 → 𝑌 Deferrer 𝐷: 𝑋 → 0,1 𝑚 . . If 𝐷1 = 1 If 𝐷2 = 1 If 𝐷3 = 1 If 𝐷4 = 1 If 𝐷𝑚−1 = 1 If 𝐷𝑚 = 1 Deferrer 𝐷 choose a committee of experts. The majority decision of committee is the final prediction Final output of selected committee Experts might have access to additional information, including group membership 𝑍. There might be a cost/penalty associated with each expert review.
Fairness in human-in-the-loop settings • Joint learning framework to learn a classifier and a deferral system for multiple experts simultaneously • Synthetic and real-world experiments on the efficacy of our method V. Keswani, M. Lease, K. Kenthapadi, Towards Unbiased and Accurate Deferral to Multiple Experts, AIES 2021.
Minimax Group Fairness: Algorithms and Experiments • Equality of error rates: an intuitive & well-studied group fairness notion • May require artificially inflating error on easier-to-predict groups  • Undesirable when most/all of the targeted population is disadvantaged • Goal: minimize maximum group error [Martinez et al, 2020] • “Ensure that the worst-off group is as well-off as possible” • Our work: algorithms based on a zero-sum game between a Learner and a Regulator • Theoretical results and experimental evaluation E. Diana, W. Gill, M. Kearns, K. Kenthapadi, A. Roth, Minimax Group Fairness: Algorithms and Experiments, AIES 2021.
Robust Interpretability of Neural Text Classifiers • Feature attribution methods used for understanding model predictions • How robust are feature attributions for neural text classifiers? • Are they identical under different random initializations of the same model? • Do they differ between a model with trained parameters and a model with random parameters? • Common feature attribution methods fail both tests! M. B. Zafar, M. Donini, D. Slack, C. Archambeau, S. Das, K. Kenthapadi, On the Lack of Robust Interpretability of Neural Text Classifiers, Findings in ACL 2021.
Privacy Research @ Amazon - Sampler
Differentially Private Query Release • Problem: answering marginal queries privately with high accuracy • Marginal queries are a special class of linear queries that count slices of a dataset. • “How many authors have visited Charlotte, graduated in the last two year and work in the Bay Area?” – A 3-way marginal query on the below dataset. • Privacy: Marginals computed against our dataset should protect against inferences on an individual’s membership (using Differential Privacy)
Differentially Private Query Release • Projection Mechanism: Evaluate noisy answers to all queries in a query class Q and find the synthetic dataset (D′) in the space of feasible datasets that minimizes error with respect to some norm. Q is a class of queries. Nikolov, Talwar, Zhang, The geometry of differential privacy: the sparse and approximate cases, STOC 2013
Differentially Private Query Release: Key Ideas 1. Relax the data domain: one-hot encode the non-continuous data and expand the domain to real numbers. Extend the differentiable queries to the new domain. 2. Adaptively select queries: repeatedly choose the k worst performing queries privately and optimize D′ to answer those well. S. Aydore, W. Brown, M. Kearns, K. Kenthapadi, L. Melis, A. Roth, A. Siva, Differentially Private Query Release Through Adaptive Projection, ICML 2021.
Responsible AI Case Studies at Google Ben Packer Google AI
Google Assistant
Google Assistant Key Points: • Think about user harms How does your product make people feel • Adversarial ("stress") testing for all Google Assistant launches • People might say racist, sexist, homophobic stuff • Diverse testers • Think about expanding who your users could and should be • Consider the diversity of your users
Computer Vision
Google Camera Key points: • Check for unconscious bias • Comprehensive testing: "make sure this works for everybody"
Night Sight
This is a “Shirley Card” Named after a Kodak studio model named Shirley Page, they were the primary method for calibrating color when processing film. SKIN TONE IN PHOTOGRAPHY SOURCES Color film was built for white people. Here's what it did to dark skin. (Vox) How Kodak's Shirley Cards Set Photography's Skin-Tone Standard, NPR
Until about 1990, virtually all Shirley Cards featured Caucasian women. SKIN TONE IN PHOTOGRAPHY SOURCES Color film was built for white people. Here's what it did to dark skin. (Vox) Colour Balance, Image Technologies, and Cognitive Equity, Roth How Photography Was Optimized for White Skin Color (Priceonomics)
As a result, photos featuring people with light skin looked fairly accurate. SKIN TONE IN PHOTOGRAPHY SOURCES Color film was built for white people. Here's what it did to dark skin. (Vox) Colour Balance, Image Technologies, and Cognitive Equity, Roth How Photography Was Optimized for White Skin Color (Priceonomics) Film Kodachrome Year 1970 Credit Darren Davis, Flickr
Photos featuring people with darker skin, not so much... SKIN TONE IN PHOTOGRAPHY SOURCES Color film was built for white people. Here's what it did to dark skin. (Vox) Colour Balance, Image Technologies, and Cognitive Equity, Roth How Photography Was Optimized for White Skin Color (Priceonomics) Film Kodachrome Year 1958 Credit Peter Roome, Flickr
Google Clips
Google Clips "We created controlled datasets by sampling subjects from different genders and skin tones in a balanced manner, while keeping variables like content type, duration, and environmental conditions constant. We then used this dataset to test that our algorithms had similar performance when applied to different groups." https://ai.googleblog.com/2018/05/automat ic-photography-with-google-clips.html
Geena Davis Inclusion Quotient [with Geena Davis Institute on Gender in Media]
Smart Compose
Adversarial Testing for Smart Compose in Gmail
Adversarial Testing fEmbedding Modelin Gmail https://developers.googleblog.com/2018/04/text-embedding-models-contain-bias.html
Adversarial Testing fEmbedding Modelin Gmail https://developers.googleblog.com/2018/04/text-embedding-models-contain-bias.html
Adversarial Testing for Smart Compose in Gmail
Machine Translation
(Historical) Gender Pronouns in Translate
Three Step Approach
1. Detect Gender-Neutral Queries Train a text classifier to detect when a Turkish query is gender-neutral. • trained on thousands of human-rated Turkish examples
2. Generate Gender-Specific Translations • Training: Modify training data to add an additional input token specifying the required gender: • (<2MALE> O bir doktor, He is a doctor) • (<2FEMALE> O bir doktor, She is a doctor) • Deployment: If step (1) predicted query is gender-neutral, add male and female tokens to query • O bir doktor -> {<2MALE> O bir doktor, <2FEMALE> O bir doktor}
3. Check for Accuracy Verify: 1. If the requested feminine translation is feminine. 2. If the requested masculine translation is masculine. 3. If the feminine and masculine translations are exactly equivalent with the exception of gender-related changes.
Result: Reduced Gender Bias in Translate
Key Takeaways Krishnaram Kenthapadi Amazon AWS AI
Good ML Practices Go a Long Way Lots of low hanging fruit in terms of improving fairness simply by using machine learning best practices • Representative data • Introspection tools • Visualization tools • Testing 01 Fairness improvements often lead to overall improvements • It’s a common misconception that it’s always a tradeoff 02
Breadth and Depth Required Looking End-to-End is critical • Need to be aware of bias and potential problems at every stage of product and ML pipelines (from design, data gathering, … to deployment and monitoring) 01 Details Matter • Slight changes in features or labeler criteria can change the outcome • Must have experts who understand the effects of decisions • Many details are not technical such as how labelers are hired 02
Process Best Practices Identify product goals Get the right people in the room Identify stakeholders Select a fairness approach Analyze and evaluate your system Mitigate issues Monitor Continuously and Escalation Plans Auditing and Transparency Policy Technology
Beyond Accuracy Performance and Cost Fairness and Bias Transparency and Explainability Privacy Security Safety Robustness
Fairness, Explainability & Privacy: Opportunities
Fairness in ML Application specific challenges Conversational AI systems: Unique bias/fairness/ethics considerations E.g., Hate speech, Complex failure modes Beyond protected categories, e.g., accent, dialect Entire ecosystem (e.g., including apps such as Alexa skills) Two-sided markets: e.g., fairness to buyers and to sellers, or to content consumers and producers Fairness in advertising (externalities) Tools for ensuring fairness (measuring & mitigating bias) in AI lifecycle Pre-processing (representative datasets; modifying features/labels) ML model training with fairness constraints Post-processing Experimentation & Post-deployment
Key Open Problems in Applied Fairness What if you don’t have the sensitive attributes? When should you use what approach? For example, Equal treatment vs equal outcome? How to identify harms? Process for framing AI problems: Will the chosen metrics lead to desired results? How to tell if data generation and collection method is appropriate for a task? (e.g., causal structure analysis?) Processes for mitigating harms and misbehaviors quickly
Explainability in ML Actionable explanations Balance between explanations & model secrecy Robustness of explanations to failure modes (Interaction between ML components) Application-specific challenges Conversational AI systems: contextual explanations Gradation of explanations Tools for explanations across AI lifecycle Pre & post-deployment for ML models Model developer vs. End user focused
Privacy in ML Privacy for highly sensitive data: model training & analytics using secure enclaves, homomorphic encryption, federated learning / on- device learning, or a hybrid Privacy-preserving model training, robust against adversarial membership inference attacks (Dynamic settings + Complex data / model pipelines) Privacy-preserving mechanisms for data marketplaces
Reflections “Fairness, Explainability, and Privacy by Design” when building AI products Collaboration/consensus across key stakeholders NYT / WSJ / ProPublica test :)
Related Tutorials / Resources • ACM Conference on Fairness, Accountability, and Transparency (ACM FAccT) • AAAI/ACM Conference on Artificial Intelligence, Ethics, and Society (AIES) • Sara Hajian, Francesco Bonchi, and Carlos Castillo, Algorithmic bias: From discrimination discovery to fairness-aware data mining, KDD Tutorial, 2016. • Solon Barocas and Moritz Hardt, Fairness in machine learning, NeurIPS Tutorial, 2017. • Kate Crawford, The Trouble with Bias, NeurIPS Keynote, 2017. • Arvind Narayanan, 21 fairness definitions and their politics, FAccT Tutorial, 2018. • Sam Corbett-Davies and Sharad Goel, Defining and Designing Fair Algorithms, Tutorials at EC 2018 and ICML 2018. • Ben Hutchinson and Margaret Mitchell, Translation Tutorial: A History of Quantitative Fairness in Testing, FAccT Tutorial, 2019. • Henriette Cramer, Kenneth Holstein, Jennifer Wortman Vaughan, Hal Daumé III, Miroslav Dudík, Hanna Wallach, Sravana Reddy, and Jean Garcia-Gathright, Translation Tutorial: Challenges of incorporating algorithmic fairness into industry practice, FAccT Tutorial, 2019.
Related Tutorials / Resources • Sarah Bird, Ben Hutchinson, Krishnaram Kenthapadi, Emre Kiciman, Margaret Mitchell, Fairness-Aware Machine Learning: Practical Challenges and Lessons Learned, Tutorials at WSDM 2019, WWW 2019, KDD 2019. • Krishna Gade, Sahin Cem Geyik, Krishnaram Kenthapadi, Varun Mithal, Ankur Taly, Explainable AI in Industry, Tutorials at KDD 2019, FAccT 2020, WWW 2020. • Himabindu Lakkaraju, Julius Adebayo, Sameer Singh, Explaining Machine Learning Predictions: State-of-the-art, Challenges, and Opportunities, NeurIPS 2020 Tutorial. • Kamalika Chaudhuri, Anand D. Sarwate, Differentially Private Machine Learning: Theory, Algorithms, and Applications, NeurIPS 2017 Tutorial. • Krishnaram Kenthapadi, Ilya Mironov, Abhradeep Guha Thakurta, Privacy-preserving Data Mining in Industry, Tutorials at KDD 2018, WSDM 2019, WWW 2019.
Thanks! Questions? •Tutorial website: https://sites.google.com/view/ResponsibleAITutorial •Feedback most welcome  • kenthk@amazon.com, bpacker@google.com, mehrnoosh.sameki@microsoft.com, nashlies@amazon.com

More Related Content

PPTX
Explainable AI in Industry (KDD 2019 Tutorial)
Krishnaram Kenthapadi
 
PPTX
Responsible AI in Industry (Tutorials at AAAI 2021, FAccT 2021, and WWW 2021)
Krishnaram Kenthapadi
 
PPTX
Explainable AI in Industry (WWW 2020 Tutorial)
Krishnaram Kenthapadi
 
PPTX
Explainable AI in Industry (FAT* 2020 Tutorial)
Krishnaram Kenthapadi
 
PPTX
Fairness-aware Machine Learning: Practical Challenges and Lessons Learned (WS...
Krishnaram Kenthapadi
 
PDF
Responsible AI
Data Con LA
 
PDF
Unlocking the Power of Generative AI An Executive's Guide.pdf
PremNaraindas1
 
PPTX
The Ethics of AI
Mark S. Steed
 
Explainable AI in Industry (KDD 2019 Tutorial)
Krishnaram Kenthapadi
 
Responsible AI in Industry (Tutorials at AAAI 2021, FAccT 2021, and WWW 2021)
Krishnaram Kenthapadi
 
Explainable AI in Industry (WWW 2020 Tutorial)
Krishnaram Kenthapadi
 
Explainable AI in Industry (FAT* 2020 Tutorial)
Krishnaram Kenthapadi
 
Fairness-aware Machine Learning: Practical Challenges and Lessons Learned (WS...
Krishnaram Kenthapadi
 
Responsible AI
Data Con LA
 
Unlocking the Power of Generative AI An Executive's Guide.pdf
PremNaraindas1
 
The Ethics of AI
Mark S. Steed
 

What's hot (20)

PPTX
Responsible AI
Neo4j
 
PPTX
Responsible AI in Industry: Practical Challenges and Lessons Learned
Krishnaram Kenthapadi
 
PDF
AI Governance – The Responsible Use of AI
NUS-ISS
 
PDF
“Responsible AI: Tools and Frameworks for Developing AI Solutions,” a Present...
Edge AI and Vision Alliance
 
PPTX
Generative AI Risks & Concerns
Ajitesh Kumar
 
PPTX
The Future of AI is Generative not Discriminative 5/26/2021
Steve Omohundro
 
PDF
Responsible AI
Neo4j
 
PDF
Fairness-aware Machine Learning: Practical Challenges and Lessons Learned (KD...
Krishnaram Kenthapadi
 
PDF
The Future is in Responsible Generative AI
Saeed Al Dhaheri
 
PPTX
Explainable AI in Industry (AAAI 2020 Tutorial)
Krishnaram Kenthapadi
 
PPTX
Measures and mismeasures of algorithmic fairness
Manojit Nandi
 
PPTX
Fairness in AI (DDSW 2019)
GoDataDriven
 
PPTX
Generative AI and law.pptx
Chris Marsden
 
PPTX
Fairness and Privacy in AI/ML Systems
Krishnaram Kenthapadi
 
PPTX
Responsible AI
Anand Rao
 
PDF
Explainable AI (XAI)
Manojkumar Parmar
 
PDF
Explainability and bias in AI
Bill Liu
 
PDF
Generative AI for the rest of us
Massimo Ferre'
 
PPTX
Responsible AI
Object Automation
 
PDF
"I don't trust AI": the role of explainability in responsible AI
Erika Agostinelli
 
Responsible AI
Neo4j
 
Responsible AI in Industry: Practical Challenges and Lessons Learned
Krishnaram Kenthapadi
 
AI Governance – The Responsible Use of AI
NUS-ISS
 
“Responsible AI: Tools and Frameworks for Developing AI Solutions,” a Present...
Edge AI and Vision Alliance
 
Generative AI Risks & Concerns
Ajitesh Kumar
 
The Future of AI is Generative not Discriminative 5/26/2021
Steve Omohundro
 
Responsible AI
Neo4j
 
Fairness-aware Machine Learning: Practical Challenges and Lessons Learned (KD...
Krishnaram Kenthapadi
 
The Future is in Responsible Generative AI
Saeed Al Dhaheri
 
Explainable AI in Industry (AAAI 2020 Tutorial)
Krishnaram Kenthapadi
 
Measures and mismeasures of algorithmic fairness
Manojit Nandi
 
Fairness in AI (DDSW 2019)
GoDataDriven
 
Generative AI and law.pptx
Chris Marsden
 
Fairness and Privacy in AI/ML Systems
Krishnaram Kenthapadi
 
Responsible AI
Anand Rao
 
Explainable AI (XAI)
Manojkumar Parmar
 
Explainability and bias in AI
Bill Liu
 
Generative AI for the rest of us
Massimo Ferre'
 
Responsible AI
Object Automation
 
"I don't trust AI": the role of explainability in responsible AI
Erika Agostinelli
 
Ad

Similar to Responsible AI in Industry (ICML 2021 Tutorial) (20)

PPTX
Responsible AI in Industry: Practical Challenges and Lessons Learned
Krishnaram Kenthapadi
 
PPTX
Fairness and Privacy in AI/ML Systems
Krishnaram Kenthapadi
 
PPTX
Investigating Fairness of Decision Making
Steffen Staab
 
PDF
Spark + AI Summit - The Importance of Model Fairness and Interpretability in ...
Francesca Lazzeri, PhD
 
PDF
Practical Applications of Machine Learning in Cybersecurity
scoopnewsgroup
 
PDF
Responsible Data Use in AI - core tech pillars
Sofus Macskássy
 
PDF
Trusted, Transparent and Fair AI using Open Source
Animesh Singh
 
PDF
Ethical Considerations in AI Development- Ensuring Fairness and Transparency
Arpan Buwa
 
PDF
Model bias in AI
Jason Tamara Widjaja
 
PPTX
Explainability for Natural Language Processing
Yunyao Li
 
PDF
How do we train AI to be Ethical and Unbiased?
Mark Borg
 
PDF
When the AIs failures send us back to our own societal biases
Clément DUFFAU
 
PPTX
[DSC Adria 23] Muthu Ramachandran AI Ethics Framework for Generative AI such ...
DataScienceConferenc1
 
PPTX
Ethical AI - Open Compliance Summit 2020
Debmalya Biswas
 
PDF
Explainability for Natural Language Processing
Yunyao Li
 
PDF
AI Developments and Trends (OECD)
AnandSRao1962
 
PPTX
Artificial Intelligence (AI) and Bias.pptx
Dr.A.Prabaharan Professor & Research Director, Public Action
 
PPTX
Machine Learning: Addressing the Disillusionment to Bring Actual Business Ben...
Jon Mead
 
PPTX
Explainability for Natural Language Processing
Yunyao Li
 
PDF
KnowMe and ShareMe: understanding automatically discovered personality traits...
Leon Gou
 
Responsible AI in Industry: Practical Challenges and Lessons Learned
Krishnaram Kenthapadi
 
Fairness and Privacy in AI/ML Systems
Krishnaram Kenthapadi
 
Investigating Fairness of Decision Making
Steffen Staab
 
Spark + AI Summit - The Importance of Model Fairness and Interpretability in ...
Francesca Lazzeri, PhD
 
Practical Applications of Machine Learning in Cybersecurity
scoopnewsgroup
 
Responsible Data Use in AI - core tech pillars
Sofus Macskássy
 
Trusted, Transparent and Fair AI using Open Source
Animesh Singh
 
Ethical Considerations in AI Development- Ensuring Fairness and Transparency
Arpan Buwa
 
Model bias in AI
Jason Tamara Widjaja
 
Explainability for Natural Language Processing
Yunyao Li
 
How do we train AI to be Ethical and Unbiased?
Mark Borg
 
When the AIs failures send us back to our own societal biases
Clément DUFFAU
 
[DSC Adria 23] Muthu Ramachandran AI Ethics Framework for Generative AI such ...
DataScienceConferenc1
 
Ethical AI - Open Compliance Summit 2020
Debmalya Biswas
 
Explainability for Natural Language Processing
Yunyao Li
 
AI Developments and Trends (OECD)
AnandSRao1962
 
Artificial Intelligence (AI) and Bias.pptx
Dr.A.Prabaharan Professor & Research Director, Public Action
 
Machine Learning: Addressing the Disillusionment to Bring Actual Business Ben...
Jon Mead
 
Explainability for Natural Language Processing
Yunyao Li
 
KnowMe and ShareMe: understanding automatically discovered personality traits...
Leon Gou
 
Ad

More from Krishnaram Kenthapadi (8)

PPTX
Amazon SageMaker Clarify
Krishnaram Kenthapadi
 
PPTX
Privacy in AI/ML Systems: Practical Challenges and Lessons Learned
Krishnaram Kenthapadi
 
PPTX
Fairness-aware Machine Learning: Practical Challenges and Lessons Learned (WW...
Krishnaram Kenthapadi
 
PPTX
Privacy-preserving Data Mining in Industry (WWW 2019 Tutorial)
Krishnaram Kenthapadi
 
PPTX
Privacy-preserving Data Mining in Industry (WSDM 2019 Tutorial)
Krishnaram Kenthapadi
 
PPTX
Fairness, Transparency, and Privacy in AI @ LinkedIn
Krishnaram Kenthapadi
 
PPTX
Privacy-preserving Analytics and Data Mining at LinkedIn
Krishnaram Kenthapadi
 
PPTX
Privacy-preserving Data Mining in Industry: Practical Challenges and Lessons ...
Krishnaram Kenthapadi
 
Amazon SageMaker Clarify
Krishnaram Kenthapadi
 
Privacy in AI/ML Systems: Practical Challenges and Lessons Learned
Krishnaram Kenthapadi
 
Fairness-aware Machine Learning: Practical Challenges and Lessons Learned (WW...
Krishnaram Kenthapadi
 
Privacy-preserving Data Mining in Industry (WWW 2019 Tutorial)
Krishnaram Kenthapadi
 
Privacy-preserving Data Mining in Industry (WSDM 2019 Tutorial)
Krishnaram Kenthapadi
 
Fairness, Transparency, and Privacy in AI @ LinkedIn
Krishnaram Kenthapadi
 
Privacy-preserving Analytics and Data Mining at LinkedIn
Krishnaram Kenthapadi
 
Privacy-preserving Data Mining in Industry: Practical Challenges and Lessons ...
Krishnaram Kenthapadi
 

Recently uploaded (20)

PPTX
LESSON-2-Roles-of-ICT-in-Teaching-for-learning_123922 (1).pptx
renavieramopiquero
 
PPTX
原版北不列颠哥伦比亚大学毕业证文凭UNBC成绩单2025年新版在线制作学位证书
e7nw4o4
 
PPTX
How tech helps people in the modern era.
upadhyayaryan154
 
PDF
LB# 820-1889_051-7370_C000.schematic.pdf
matheusalbuquerqueco3
 
PPTX
Artificial-Intelligence-in-Daily-Life (2).pptx
nidhigoswami335
 
PDF
DNSSEC Made Easy, presented at PHNOG 2025
APNIC
 
PDF
PDF document: World Game (s) Great Redesign.pdf
Steven McGee
 
PDF
Slides: PDF Eco Economic Epochs for World Game (s) pdf
Steven McGee
 
PDF
APNIC Update, presented at PHNOG 2025 by Shane Hermoso
APNIC
 
PPTX
dns domain name system history work.pptx
MUHAMMADKAVISHSHABAN
 
PDF
UI/UX Developer Guide: Tools, Trends, and Tips for 2025
Penguin peak
 
PPTX
Different Generation Of Computers .pptx
divcoder9507
 
PPTX
办理方法西班牙假毕业证蒙德拉贡大学成绩单MULetter文凭样本
xxxihn4u
 
PPTX
The Internet of Things (IoT) refers to a vast network of interconnected devic...
chethana8182
 
PDF
Data Protection & Resilience in Focus.pdf
AmyPoblete3
 
PPTX
Blue and Dark Blue Modern Technology Presentation.pptx
ap177979
 
PPTX
Perkembangan Perangkat jaringan komputer dan telekomunikasi 3.pptx
Prayudha3
 
PPTX
Pengenalan perangkat Jaringan komputer pada teknik jaringan komputer dan tele...
Prayudha3
 
PDF
The Internet of Things (IoT) refers to a vast network of interconnected devic...
chethana8182
 
PPTX
Google SGE SEO: 5 Critical Changes That Could Wreck Your Rankings in 2025
Reversed Out Creative
 
LESSON-2-Roles-of-ICT-in-Teaching-for-learning_123922 (1).pptx
renavieramopiquero
 
原版北不列颠哥伦比亚大学毕业证文凭UNBC成绩单2025年新版在线制作学位证书
e7nw4o4
 
How tech helps people in the modern era.
upadhyayaryan154
 
LB# 820-1889_051-7370_C000.schematic.pdf
matheusalbuquerqueco3
 
Artificial-Intelligence-in-Daily-Life (2).pptx
nidhigoswami335
 
DNSSEC Made Easy, presented at PHNOG 2025
APNIC
 
PDF document: World Game (s) Great Redesign.pdf
Steven McGee
 
Slides: PDF Eco Economic Epochs for World Game (s) pdf
Steven McGee
 
APNIC Update, presented at PHNOG 2025 by Shane Hermoso
APNIC
 
dns domain name system history work.pptx
MUHAMMADKAVISHSHABAN
 
UI/UX Developer Guide: Tools, Trends, and Tips for 2025
Penguin peak
 
Different Generation Of Computers .pptx
divcoder9507
 
办理方法西班牙假毕业证蒙德拉贡大学成绩单MULetter文凭样本
xxxihn4u
 
The Internet of Things (IoT) refers to a vast network of interconnected devic...
chethana8182
 
Data Protection & Resilience in Focus.pdf
AmyPoblete3
 
Blue and Dark Blue Modern Technology Presentation.pptx
ap177979
 
Perkembangan Perangkat jaringan komputer dan telekomunikasi 3.pptx
Prayudha3
 
Pengenalan perangkat Jaringan komputer pada teknik jaringan komputer dan tele...
Prayudha3
 
The Internet of Things (IoT) refers to a vast network of interconnected devic...
chethana8182
 
Google SGE SEO: 5 Critical Changes That Could Wreck Your Rankings in 2025
Reversed Out Creative
 

Responsible AI in Industry (ICML 2021 Tutorial)

  • 1. Responsible AI in Industry: Practical Challenges and Lessons Learned TutoriaI 2021 Krishnaram Kenthapadi (Amazon AWS AI), Ben Packer (Google AI), Mehrnoosh Sameki (Microsoft Azure), Nashlie Sephus (Amazon AWS AI) https://sites.google.com/view/ResponsibleAITutorial
  • 2. Massachusetts Group Insurance Commission (1997): Anonymized medical history of state employees William Weld vs Latanya Sweeney Latanya Sweeney (MIT grad student): $20 – Cambridge voter roll born July 31, 1945 resident of 02138
  • 3. 64% Uniquely identifiable with ZIP + birth date + gender (in the US population) Golle, “Revisiting the Uniqueness of Simple Demographics in the US Population”, WPES 2006
  • 4. A History of Privacy Failures … Credit: Kobbi Nissim,Or Sheffet
  • 7. • Ethical challenges posed by AI systems • Inherent biases present in society • Reflected in training data • AI/ML models prone to amplifying such biases Algorithmic Bias
  • 8. Laws against Discrimination Immigration Reform and Control Act Citizenship Rehabilitation Act of 1973; Americans with Disabilities Act of 1990 Disability status Civil Rights Act of 1964 Race Age Discrimination in Employment Act of 1967 Age Equal Pay Act of 1963; Civil Rights Act of 1964 Sex And more...
  • 10. Motivation & Business Opportunities Regulatory. We need to understand why the ML model made a given prediction and also whether the prediction it made was free from bias, both in training and at inference. Business. Providing explanations to internal teams (loan officers, customer service rep, compliance teams) and end users/customers Data Science. Improving models through better feature engineering and training data generation, understanding failure modes of the model, debugging model predictions, etc.
  • 11. © 2020, Amazon Web Services, Inc. or its Affiliates. Amazon SageMaker VISION SPEECH TEXT SEARCH CHATBOTS PERSONALIZATION FORECASTING FRAUD CONTACT CENTERS Deep Learning AMIs & Containers GPUs & CPUs Elastic Inference Trainium Inferentia FPGA AI SERVICES ML SERVICES FRAMEWORKS & INFRASTRUCTURE DeepGraphLibrary Amazon Rekognition Amazon Polly Amazon Transcribe +Medical Amazon Lex Amazon Personalize Amazon Forecast Amazon Comprehend +Medical Amazon Textract Amazon Kendra Amazon CodeGuru Amazon Fraud Detector Amazon Translate INDUSTRIAL AI CODE AND DEVOPS NEW Amazon DevOps Guru Voice ID For AmazonConnect Contact Lens NEW Amazon Monitron NEW AWS Panorama + Appliance NEW Amazon Lookout for Vision NEW Amazon Lookout for Equipment Scaling Fairness, Explainability & Privacy across the AWS ML Stack NEW Amazon HealthLake HEALTHCARE AI NEW Amazon Lookout for Metrics ANOMALY DETECTION Amazon Transcribe for Medical Amazon Comprehend for Medical Label data NEW Aggregate & prepare data NEW Store & share features Auto ML Spark/R NEW Detect bias Visualize in notebooks Pick algorithm Train models Tune parameters NEW Debug & profile Deploy in production Manage & monitor NEW CI/CD Human review review NEW: Model management for edge devices NEW: SageMaker JumpStart SAGEMAKER STUDIO IDE
  • 12. LinkedIn operates the largest professional network on the Internet Tell your story 740M members 55M+ companies are represented on LinkedIn 90K schools listed (high school & college) 36K skills listed 14M+ open jobs on LinkedIn Jobs 280B Feed updates
  • 13. Tutorial Outline • Fairness-aware ML: An overview • Explainable AI: An overview • Privacy-preserving ML: An overview • Responsible AI tools • Case studies • Key takeaways
  • 14. Fairness-aware ML: An Overview Nashlie Sephus, PhD Applied Science Manager, AWS AI
  • 15. • ML Fairness Considerations • ML and Humans • What is fairness/inclusion for ML? • Where May Biases Occur? • Testing Techniques with Face Experiments • Takeaways Outline 15
  • 16. Product Introspection (1): Make Your Key Choices Explicit [Mitchell et al., 2018] Goals Decision Prediction Profit from loans Whether to lend Loan will be repaid Justice, Public safety Whether to detain Crime committed if not detained • Goals are ideally measurable • What are your non-goals? • Which decisions are you not considering? • What is the relationship between Prediction and Decision?
  • 17. Product Introspection (2): Identify Potential Harms • What are the potential harms? • Applicants who would have repaid are not given loans • Convicts who would not commit a crime are locked up. • Are there also longer term harms? • Applicants are given loans, then go on to default, harming their credit score • Are some harms especially bad?
  • 18. Seek out Diverse Perspectives • Fairness Experts • User Researchers • Privacy Experts • Legal • Social Science Backgrounds • Diverse Identities • Gender • Sexual Orientation • Race • Nationality • Religion
  • 20. ! 16
  • 22. 22
  • 23. Error-free (no system is perfect) 100% confident Intended to replace human judgement What ML Is Not 23
  • 27. Detect presence of a face in an image or a video. Face Detection 27
  • 28. A system to determine the gender, age, emotion, presence of facial hair, etc. from a detected face. Face Analysis
  • 29. A system to determine a detected faces identity by matching it against a database of faces and their associated identities. Face Recognition 29
  • 30. Estimation of the confidence or certainty of any prediction Expressed in the form of a probability or confidence score Confidence Score 30
  • 31. Face Recognition: Common Causes of Errors ILLUMINATION VARIANCE POSE / VIEWPOINT AGING EXPRESSION / STYLE OCCLUSION Lighting, camera controls like exposure, shadows, highlights Face pose, camera angles Natural aging, artificial makeup Face expression like laughing, facial hair such as a beard, hair style Part of the face hidden as in group pictures 31
  • 32. Where Can Biases Exist? 32
  • 33. AFRICAN SCANDINAVIAN A B C The PPB dataset [Buolamwini 2018] 6.3% 20.8% GenderShades.Org 33 [Buolamwini & Gebru 2018]
  • 35. Racial Comparisons of Datasets [FairFace] 35
  • 40. Launch with Confidence: Testing for Bias • How will you know if users are being harmed? • How will you know if harms are unfairly distributed? • Detailed testing practices are often not covered in academic papers • Discussing testing requirements is a useful focal point for cross-functional teams
  • 41. Reproducibility - Notebook Experiments 41
  • 43. 43
  • 45. 45
  • 46. 46
  • 47. 47
  • 48. 48
  • 50. Gender Classification w.r.t. Hair Lengths – PPB2 50
  • 53. Model Predictions Evaluate for Inclusion - Confusion Matrix
  • 54. Model Predictions Positive Negative Evaluate for Inclusion - Confusion Matrix
  • 55. Model Predictions Positive Negative ● Exists ● Predicted True Positives ● Doesn’t exist ● Not predicted True Negatives Evaluate for Inclusion - Confusion Matrix
  • 56. Model Predictions Positive Negative ● Exists ● Predicted True Positives ● Exists ● Not predicted False Negatives ● Doesn’t exist ● Predicted False Positives ● Doesn’t exist ● Not predicted True Negatives Evaluate for Inclusion - Confusion Matrix
  • 57. Efficient Testing for Bias • Development teams are under multiple constraints • Time • Money • Human resources • Access to data • How can we efficiently test for bias? • Prioritization • Strategic testing
  • 58. Choose your evaluation metrics in light of acceptable tradeoffs between False Positives and False Negatives
  • 59. Takeaways • Testing for blindspots amongst intersectionality is key. • Taking into account confidence scores/thresholds and error bars when measuring for biases is necessary. • Representation matters. • Transparency, reproducibility, and education can promote change. • Confidence in your product's fairness requires fairness testing • Fairness testing has a role throughout the product iteration lifecycle • Contextual concerns should be used to prioritize fairness testing 59
  • 60. AI Fairness and Transparency Tools Mehrnoosh Sameki Microsoft Azure
  • 64. Understand and debug your model Glassbox Models: Model Types: Linear Models, Decision Trees, Decision Rules, Explainable Boosting Machines Blackbox Models: Model Formats: Python models using scikit predict convention, Scikit, Tensorflow, Pytorch, Keras, Explainers: SHAP, LIME, Global Surrogate, Feature Permutation DiCE Diverse Counterfactual Explanations Interpret Glassbox and blackbox interpretability methods for tabular data Interpret-text Interpretability methods for text data Interpret-community Community-driven interpretability techniques for tabular data https://github.com/interpretml Azureml-interpret AzureML SDK wrapper for Interpret and Interpret-community
  • 66. Models designed to be interpretable. Lossless explainability. Fever? Internal Bleeding? Stay Home Stay Home Go to Hospital Decision Trees Rule Lists Linear Models ….
  • 67. Explain any ML system. Approximate explainability. Model Explanation Perturb Inputs Analyze SHAP LIME Partial Dependence Sensitivity Analysis
  • 68. Fairness Useful links: • AI Show • Tutorial Video • Customer Highlight
  • 69. There are many ways that an AI system can behave unfairly. Fairness in AI Avoiding negative outcomes of AI systems for different groups of people A model for screening loan or job application might be much better at picking good candidates among white men than among other groups. A voice recognition system might fail to work as well for women as it does for men.
  • 70. https://github.com/fairlearn/fairlearn Assessing unfairness in your model Fairness Assessment: Usecommonfairness metricsandaninteractive dashboardto assess which groups of peoplemay benegatively impacted. Model Formats: Python models using scikit predict convention, Scikit, Tensorflow, Pytorch, Keras Metrics: 15+ Common group fairness metrics Model Types: Classification, Regression Unfairness Mitigation: Use state-of-the-art algorithmsto mitigate unfairness in yourclassificationandregressionmodels.
  • 71. Fairness Assessment Input Selections Sensitive attribute Performance metric Assessment Results Disparity in performance Disparity in predictions Mitigation Algorithms Post-processing algorithm Reductions Algorithm
  • 72. Situation: Solution: Philips Healthcare used Fairlearn to check whether our ICU models perform similarly for patients with different ethnicities and gender identities, etc. Microsoft CSE (Led by Tempest Van Schaik) collaborated with Philips to build a solution using Azure DevOps pipelines, Azure Databricks and Mlflow. Built a pipeline to make fairness monitoring routine, checking the fairness of predictions for patients of different genders, ethnicities, and medical conditions, using Fairlearn metrics. Fairness analysis helped show that Philips’ predictive model performs better than industry standard ICU models Standard model predictions for a patient differ depending on how the ICU documented their test results. Customer: Philips Industry: Healthcare Size: 80,000+ employees Country: Netherlands Products and services: Microsoft Azure DevOps Microsoft Azure Databricks MLFlow Putting fairness monitoring in production with ICU models Philips Healthcare Informatics Philips Healthcare Informatics helps ICUs benchmark their performance (e.g. mortality rate). They create quarterly benchmark reports that compare actual performance vs performance predicted by ML models. They have models trained on the largest ICU dataset in USA: 400+ ICUs, 6M+ patient stays, billions of vital signs & lab tests. Deploying ICU models responsibly Philips needed a scalable, reliable, repeatable and responsible way to bring ML models into production.
  • 73. Situation: Solution: Impact: “Azure Machine Learning and its Fairlearn capabilities offer advanced fairness and explainability that have helped us deploy trustworthy AI solutions for our customers, while enabling stakeholder confidence and regulatory compliance.” —Alex Mohelsky, Partner and Advisory Data, Analytic, and AI Leader, EY Canada Customer: EY Industry: Partner Professional Services Size: 10,000+ employees Country: United Kingdom Products and services: Microsoft Azure Microsoft Azure Machine Learning Read full story here Organizations won’t fully embrace AI until they trust it. EY wanted to help its customers embrace AI to help them better understand their customers, identify fraud and security breaches sooner, and make loan decisions faster and more efficiently. The company developed its EY Trusted AI Platform, which uses Microsoft Azure Machine Learning capabilities to assess and mitigate unfairness in machine learning models. Running on Azure, the platform uses Fairlearn and InterpretML, open-source capabilities in Azure Machine Learning. When EY tested Fairlearn with real mortgage data, it reduced the accuracy disparity between men and women approved or denied loans from 7 percent to less than 0.5 percent. Through this platform, EY helps customers and regulators alike gain confidence in AI and machine learning.
  • 74. TECH SOCIETY Reception & Adoption Educational materials are key for adoption of fairness toolkits  manuals, case studies, white papers [Lee & Singh, 2020]
  • 75. • Interpretability at training time • Combination of glass-box models and black-box explainers • Auto reason code generation for local predictions • Ability to cross reference to other techniques to ensure stability and consistency in results H2O
  • 77. IBM Open Scale Goal: to provide AI operations team with a toolkit that allows for: • Monitoring and re-evaluating machine learning models after deployment
  • 78. IBM Open Scale Goal: to provide AI operations team with a toolkit that allows for: • Monitoring and re-evaluating machine learning models after deployment • ACCURACY • FAIRNESS • PERFORMANCE
  • 83. AI Fairness 360 Datasets Toolbox Fairness metrics (30+) Bias mitigation algorithms (9+) Guidance Industry-specific tutorials
  • 84. AI Fairness 360 Datasets Toolbox Fairness metrics (30+) Bias mitigation algorithms (9+) Guidance Industry-specific tutorials Pre-processing algorithm: a bias mitigation algorithm that is applied to training dat In-processing algorithm: a bias mitigation algorithm that is applied to a model during its training Post-processing algorithm: a bias mitigation algorithm that is applied to predicted labels
  • 85. What If Tool Goal: Code-free probing of machine learning models • Feature perturbations (what if scenarios) • Counterfactual example analysis • [Classification] Explore the effects of different classification thresholds, taking into account constraints such as different numerical fairness metrics.
  • 89. Datasheets for Datasets [Gebru et al., 2018] • Better data-related documentation • Datasheets for datasets: every dataset, model, or pre-trained API should be accompanied by a data sheet that documents its • Creation • Intended uses • Limitations • Maintenance • Legal and ethical considerations • Etc.
  • 90. Model Cards for Model Reporting[Mitchell et al., 2018]
  • 91. Fact Sheets [Arnold et al., 2019] • Is distinguished from “model cards” and “datasheets” in that the focus is on the final AI service: • What is the intended use of the service output? • What algorithms or techniques does this service implement? • Which datasets was the service tested on? (Provide links to datasets that were used for testing, along with corresponding datasheets.) • Describe the testing methodology. • Describe the test results. • Etc.
  • 92. Responsible AI Case Studies at LinkedIn Krishnaram Kenthapadi Amazon AWS AI
  • 93. Fairness in AI @ LinkedIn Fairness-aware Talent Search Ranking* *Work done while at LinkedIn
  • 95. Insights to Identify Diverse Talent Pools Representative Talent Search Results Diversity Learning Curriculum “Diversity by Design” in LinkedIn’s Talent Solutions
  • 97. Representative Ranking for Talent Search S. C. Geyik, S. Ambler, K. Kenthapadi, Fairness- Aware Ranking in Search & Recommendation Systems with Application to LinkedIn Talent Search, KDD’19. [Microsoft’s AI/ML conference (MLADS’18). Distinguished Contribution Award] Building Representative Talent Search at LinkedIn (LinkedIn engineering blog)
  • 98. Intuition for Measuring and Achieving Representativeness Ideal: Top ranked results should follow a desired distribution on gender/age/… E.g., same distribution as the underlying talent pool Inspired by “Equal Opportunity” definition [Hardt et al, NeurIPS’16] Defined measures (skew, divergence) based on this intuition
  • 99. Measuring (Lack of) Representativeness Skew@k (Logarithmic) ratio of the proportion of candidates having a given attribute value among the top k ranked results to the corresponding desired proportion Variants: MinSkew: Minimum over all attribute values MaxSkew: Maximum over all attribute values Normalized Discounted Cumulative Skew Normalized Discounted Cumulative KL-divergence
  • 100. Fairness-aware Reranking Algorithm (Simplified) Partition the set of potential candidates into different buckets for each attribute value Rank the candidates in each bucket according to the scores assigned by the machine-learned model Merge the ranked lists, balancing the representation requirements and the selection of highest scored candidates Representation requirement: Desired distribution on gender/age/… Algorithmic variants based on how we choose the next attribute
  • 102. Validating Our Approach Gender Representativeness Over 95% of all searches are representative compared to the qualified population of the search Business Metrics A/B test over LinkedIn Recruiter users for two weeks No significant change in business metrics (e.g., # InMails sent or accepted) Ramped to 100% of LinkedIn Recruiter users worldwide
  • 103. Lessons learned • Post-processing approach desirable • Model agnostic • Scalable across different model choices for our application • Acts as a “fail-safe” • Robust to application-specific business logic • Easier to incorporate as part of existing systems • Build a stand-alone service or component for post-processing • No significant modifications to the existing components • Complementary to efforts to reduce bias from training data & during model training • Collaboration/consensus across key
  • 104. Lessons learned • Post-processing approach desirable • Model agnostic • Scalable across different model choices for our application • Acts as a “fail-safe” • Robust to application-specific business logic • Easier to incorporate as part of existing systems • Build a stand-alone service or component for post-processing • No significant modifications to the existing components • Complementary to efforts to reduce bias from training data & during model training • Collaboration/consensus across key
  • 105. Evaluating Fairness Using Permutations Tests [DiCiccio, Vasudevan, Basu, Kenthapadi, Agarwal, KDD’20] • Is the measured discrepancy across different groups statistically significant? • Use statistical hypothesis tests! • Can we perform hypothesis tests in a metric-agnostic manner? • Non-parametric tests can help! • Permutation testing framework
  • 106. Brief Review of Permutation Tests Observe data from two populations: and Are the populations the same? A reasonable test statistic might be
  • 107. Brief Review of Permutation Tests (Continued) A p-value is the chance of observing a test statistic at least as “extreme” as the value we actually observed Permutation test approach: ●Randomly shuffle the population designations of the observations ●Recompute the test statistic T ●Repeat many times Permutation p-value: the proportion of permuted datasets resulting in a larger test statistic than the original value This test is exact!
  • 108. A Fairness Example Consider testing whether the true positive rate of a classifier is equal between two groups Test Statistic: difference in proportion of negative labeled observations that are classified as positive between the two groups Permutation test: Randomly reshuffle group labels, recompute test statistic
  • 109. Permutations Tests for Evaluating Fairness in ML Models • Issues with classical permutation test • Want to check: just equality of the fairness metric (e.g., false positive rate) across groups, and not if the two groups have identical distribution • Exact for the strong null hypothesis … • … but may not be valid (even asymptotically) for the weak null hypothesis • Our paper: A fix for this issue • Choose a pivotal statistic (asymptotically distribution-free; does not depend on the observed data’s distribution) • E.g., Studentize the test statistic
  • 110. Permutations Tests for Evaluating Fairness in ML Models • Issues with classical permutation test • Want to check: just equality of the fairness metric (e.g., false positive rate) across groups, and not if the two groups have identical distribution • Exact for the strong null hypothesis … • … but may not be valid (even asymptotically) for the weak null hypothesis • Our paper: A fix for this issue • Choose a pivotal statistic (asymptotically distribution-free; does not depend on the observed data’s distribution) • E.g., Studentize the test statistic
  • 111. Engineering for Fairness in AI Lifecycle S.Vasudevan, K. Kenthapadi, LiFT: A Scalable Framework for Measuring Fairness in ML Applications, CIKM’20 https://github.com/linkedin/LiFT
  • 112. LiFT System Architecture [Vasudevan & Kenthapadi, CIKM’20] •Flexibility of Use (Platform agnostic) •Ad-hoc exploratory analyses •Deployment in offline workflows •Integration with ML Frameworks •Scalability •Diverse fairness metrics •Conventional fairness metrics •Benefit metrics •Statistical tests
  • 113. LiFT System Architecture [Vasudevan & Kenthapadi, CIKM’20] •Flexibility of Use (Platform agnostic) •Ad-hoc exploratory analyses •Deployment in offline workflows •Integration with ML Frameworks •Scalability •Diverse fairness metrics •Conventional fairness metrics •Benefit metrics •Statistical tests
  • 114. Acknowledgements LinkedIn Talent Solutions Diversity team, Hire & Careers AI team, Anti-abuse AI team, Data Science Applied Research team Special thanks to Deepak Agarwal, Parvez Ahammad, Stuart Ambler, Kinjal Basu, Jenelle Bray, Erik Buchanan, Bee-Chung Chen, Fei Chen, Patrick Cheung, Gil Cottle, Cyrus DiCiccio, Patrick Driscoll, Carlos Faham, Nadia Fawaz, Priyanka Gariba, Meg Garlinghouse, Sahin Cem Geyik, Gurwinder Gulati, Rob Hallman, Sara Harrington, Joshua Hartman, Daniel Hewlett, Nicolas Kim, Rachel Kumar, Monica Lewis, Nicole Li, Heloise Logan, Stephen Lynch, Divyakumar Menghani, Varun Mithal, Arashpreet Singh Mor, Tanvi Motwani, Preetam Nandy, Lei Ni, Nitin Panjwani, Igor Perisic, Hema Raghavan, Romer Rosales, Guillaume Saint-Jacques, Badrul Sarwar, Amir Sepehri, Arun Swami, Ram Swaminathan, Grace Tang, Ketan Thakkar, Sriram Vasudevan, Janardhanan Vembunarayanan, James Verbus, Xin Wang, Hinkmond Wong, Ya Xu, Lin Yang, Yang Yang, Chenhui Zhai, Liang Zhang, Yani Zhang
  • 115. Privacy in AI @ LinkedIn PriPeARL: Framework to compute robust, privacy-preserving analytics
  • 116. Analytics & Reporting Products at LinkedIn Profile View Analytics 116 Content Analytics Ad Campaign Analytics All showing demographics of members engaging with the product
  • 117. Admit only a small # of predetermined query types Querying for the number of member actions, for a specified time period, together with the top demographic breakdowns Analytics & Reporting Products at LinkedIn
  • 118. Admit only a small # of predetermined query types Querying for the number of member actions, for a specified time period, together with the top demographic breakdowns Analytics & Reporting Products at LinkedIn E.g., Title = “Senior Director” E.g., Clicks on a given ad
  • 119. Privacy Requirements Attacker cannot infer whether a member performed an action E.g., click on an article or an ad Attacker may use auxiliary knowledge E.g., knowledge of attributes associated with the target member (say, obtained from this member’s LinkedIn profile) E.g., knowledge of all other members that performed similar action (say, by creating fake accounts)
  • 120. Possible Privacy Attacks 120 Targeting: Senior directors in US, who studied at Cornell Matches ~16k LinkedIn members → over minimum targeting threshold Demographic breakdown: Company = X May match exactly one person → can determine whether the person clicks on the ad or not Require minimum reporting threshold Attacker could create fake profiles! E.g. if threshold is 10, create 9 fake profiles that all click. Rounding mechanism E.g., report incremental of 10 Still amenable to attacks E.g. using incremental counts over time to infer individuals’ actions Need rigorous techniques to preserve member privacy (not reveal exact aggregate counts)
  • 121. Problem Statement Compute robust, reliable analytics in a privacy- preserving manner, while addressing the product needs.
  • 122. PriPeARL: A Framework for Privacy-Preserving Analytics K. Kenthapadi, T. T. L. Tran, ACM CIKM 2018 122 Pseudo-random noise generation, inspired by differential privacy ● Entity id (e.g., ad creative/campaign/account) ● Demographic dimension ● Stat type (impressions, clicks) ● Time range ● Fixed secret seed Uniformly Random Fraction ● Cryptographic hash ● Normalize to (0,1) Random Noise Laplace Noise ● Fixed ε True Count Noisy Count To satisfy consistency requirements ● Pseudo-random noise → same query has same result over time, avoid averaging attack. ● For non-canonical queries (e.g., time ranges, aggregate multiple entities) ○ Use the hierarchy and partition into canonical queries ○ Compute noise for each canonical queries and sum up the noisy counts
  • 124. Lessons Learned from Deployment (> 1 year) Semantic consistency vs. unbiased, unrounded noise Suppression of small counts Online computation and performance requirements Scaling across analytics applications Tools for ease of adoption (code/API library, hands-on how-to tutorial) help! Having a few entry points (all analytics apps built over Pinot)  wider adoption
  • 125. Summary Framework to compute robust, privacy-preserving analytics Addressing challenges such as preserving member privacy, product coverage, utility, and data consistency Future Utility maximization problem given constraints on the ‘privacy loss budget’ per user E.g., noise with larger variance to impressions but less noise to clicks (or conversions) E.g., more noise to broader time range sub-queries and less noise to granular time range sub-queries Reference: K. Kenthapadi, T. Tran, PriPeARL: A Framework for Privacy- Preserving Analytics and Reporting at LinkedIn, ACM CIKM 2018.
  • 126. Acknowledgements Team: AI/ML: Krishnaram Kenthapadi, Thanh T. L. Tran Ad Analytics Product & Engineering: Mark Dietz, Taylor Greason, Ian Koeppe Legal / Security: Sara Harrington, Sharon Lee, Rohit Pitke Acknowledgements Deepak Agarwal, Igor Perisic, Arun Swami
  • 128. LinkedIn Salary (launched in Nov, 2016)
  • 129. Data Privacy Challenges Minimize the risk of inferring any one individual’s compensation data Protection against data breach No single point of failure
  • 130. Problem Statement How do we design LinkedIn Salary system taking into account the unique privacy and security challenges, while addressing the product requirements? K. Kenthapadi, A. Chudhary, and S. Ambler, LinkedIn Salary: A System for Secure Collection and Presentation of Structured Compensation Insights to Job Seekers, IEEE PAC 2017 (arxiv.org/abs/1705.06976)
  • 131. Title Region $$ User Exp Designer SF Bay Area 100K User Exp Designer SF Bay Area 115K ... ... ... Title Region $$ User Exp Designer SF Bay Area 100K De-identification Example Title Region Company Industry Years of exp Degree FoS Skills $$ User Exp Designer SF Bay Area Google Internet 12 BS Interactive Media UX, Graphics, ... 100K Title Region Industry $$ User Exp Designer SF Bay Area Internet 100K Title Region Years of exp $$ User Exp Designer SF Bay Area 10+ 100K Title Region Company Years of exp $$ User Exp Designer SF Bay Area Google 10+ 100K #data points > threshold? Yes ⇒ Copy to Hadoop (HDFS) Note: Original submission stored as encrypted objects.
  • 133. Acknowledgements Team: AI/ML: Krishnaram Kenthapadi, Stuart Ambler, Xi Chen, Yiqun Liu, Parul Jain, Liang Zhang, Ganesh Venkataraman, Tim Converse, Deepak Agarwal Application Engineering: Ahsan Chudhary, Alan Yang, Alex Navasardyan, Brandyn Bennett, Hrishikesh S, Jim Tao, Juan Pablo Lomeli Diaz, Patrick Schutz, Ricky Yan, Lu Zheng, Stephanie Chou, Joseph Florencio, Santosh Kumar Kancha, Anthony Duerr Product: Ryan Sandler, Keren Baruch Other teams (UED, Marketing, BizOps, Analytics, Testing, Voice of Members, Security, …): Julie Kuang, Phil Bunge, Prateek Janardhan, Fiona Li, Bharath Shetty, Sunil Mahadeshwar, Cory Scott, Tushar Dalvi, and team Acknowledgements David Freeman, Ashish Gupta, David Hardtke, Rong Rong, Ram
  • 134. Responsible AI Case Studies at Amazon Krishnaram Kenthapadi Amazon AWS AI
  • 135. 135 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | 135 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | Detect bias in ML models and understand model predictions Amazon SageMaker Clarify
  • 136. 136 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | Predictive Maintenance Manufacturing, Automotive, IoT Demand Forecasting Retail, Consumer Goods, Manufacturing Fraud Detection Financial Services, Online Retail Credit Risk Prediction Financial Services, Retail Extract and Analyze Data from Documents Healthcare, Legal, Media/Ent, Education Computer Vision Healthcare, Pharma, Manufacturing Autonomous Driving Automotive, Transportation Personalized Recommendations Media & Entertainment, Retail, Education Churn Prediction Retail, Education, Software & Internet https://aws.amazon.c om/sagemaker/gettin g-started Amazon SageMaker Customer ML Use cases
  • 137. 137 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | Bias and Explainability: Challenges 1 Without detection, it is hard to know if bias has entered an ML model: • Imbalances may be present in the initial dataset • Bias may develop during training • Bias may develop over time after model deployment 2 Machine learning models are often complex & opaque, making explainability critical: • Regulations may require companies to be able to explain model predictions • Internal stakeholders and customers may need explanations for model behavior • Data science teams can improve models if they understand model behavior
  • 138. 138 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | Amazon SageMaker Clarify Detect bias in ML models and understand model predictions Detect bias during data preparation Identify imbalances in data Evaluate the degree to which various types of bias are present in your model Check your trained model for bias Understand the relative importance of each feature to your model’s behavior Explain overall model behavior Understand the relative importance of each feature for individual inferences Explain individual predictions Provide alerts and detect drift over time due to changing real-world conditions Detect drift in bias and model behavior over time Generated automated reports Produce reports on bias and explanations to support internal presentations
  • 139. 147 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | SageMaker Clarify Use Cases Regulatory Compliance Internal Reporting Operational Excellence Customer Service
  • 140. Lessons learned • Fairness as a Process • Notions of bias & fairness are highly application dependent • Choice of the attribute(s) for which bias is to be measured & the choice of the bias metrics to be guided by social, legal, and other non-technical considerations • Collaboration/consensus across key stakeholders • Wide spectrum of customers with different levels of technical background • Managed service vs. open source packages • Monitoring of the deployed model • Fairness & explainability considerations across the ML lifecycle
  • 141. 150 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | Fairness and Explainability by Design in the ML Lifecycle
  • 142. 151 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | Additional Pointers For more information on Amazon SageMaker Clarify, please refer: • https://aws.amazon.com/sagemaker/clarify • Amazon Science / AWS Articles • https://aws.amazon.com/blogs/aws/new-amazon-sagemaker-clarify-detects- bias-and-increases-the-transparency-of-machine-learning-models • https://www.amazon.science/latest-news/how-clarify-helps-machine-learning- developers-detect-unintended-bias • Technical papers: (1) Amazon SageMaker Clarify [KDD’21] (2) Fairness Measures for Machine Learning in Finance • https://github.com/aws/amazon-sagemaker-clarify Acknowledgments: Amazon SageMaker Clarify core team, Amazon AWS AI team, and partners across Amazon
  • 143. 152 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | Additional Pointers For more information on Amazon SageMaker Clarify, please refer: • https://aws.amazon.com/sagemaker/clarify • Amazon Science / AWS Articles • https://aws.amazon.com/blogs/aws/new-amazon-sagemaker-clarify-detects- bias-and-increases-the-transparency-of-machine-learning-models • https://www.amazon.science/latest-news/how-clarify-helps-machine-learning- developers-detect-unintended-bias • Technical papers: (1) Amazon SageMaker Clarify [KDD’21] (2) Fairness Measures for Machine Learning in Finance • https://github.com/aws/amazon-sagemaker-clarify Acknowledgments: Amazon SageMaker Clarify core team, Amazon AWS AI team, and partners across Amazon
  • 144. © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Trademark Amazon SageMaker Debugger Debug and profile ML model training and get real-time insights
  • 145. © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Trademark Why debugging and profiling Training bugs Large compute instances Long training times Training ML models is difficult and compute intensive
  • 146. © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Trademark SageMaker Debugger Real-time monitoring Relevant data capture Automatic error detection SageMaker Studio integration Debug data while training is ongoing Zero code change Persistent in your S3 bucket Built-in and custom rules Early termination Alerts about rule status Save time and cost Find issues early Accelerate prototyping Detect performance bottlenecks View suggestions on resolving bottlenecks, Interactive visualizations Monitor utilization Profile by step or time duration Right size instance Improve utilization Reduce cost System resource usage Time spent by training operations
  • 147. © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Trademark https://www.amazon.science/publications/amazon-sagemaker-debugger-a-system-for-real-time-insights-into-machine-learning-model- training
  • 148. 157 © 2020 Amazon Web Services, Inc. or its affiliates. All rights reserved | Additional Pointers For more information on Amazon SageMaker Debugger, please refer: • https://aws.amazon.com/sagemaker/debugger • AWS Articles • https://aws.amazon.com/blogs/aws/amazon-sagemaker-debugger-debug-your- machine-learning-models • https://aws.amazon.com/blogs/machine-learning/detecting-hidden-but-non-trivial- problems-in-transfer-learning-models-using-amazon-sagemaker-debugger • Technical paper: Amazon SageMaker Debugger: A System for Real-Time Insights into Machine Learning Model Training (MLSys 2021) • https://pypi.org/project/smdebug Acknowledgments: Amazon SageMaker Debugger core team, Amazon AWS AI team, and partners across Amazon
  • 149. Fairness for Opaque Models via Model Tuning (Hyperparameter Optimization) • Can we tune the hyperparameters of a model to achieve both accuracy and fairness? • Can we support both opaque models and opaque fairness constraints? • Use Bayesian optimization for HPO with fairness constraints! • Explore hyperparameter configurations where fairness constraints are satisfied V. Perrone, M. Donini, M. B. Zafar, R. Schmucker, K. Kenthapadi, C. Archambeau, Fair Bayesian Optimization, AIES 2021 (Best paper award @ ICML 2020 AutoML workshop)
  • 150. Human-in-the-loop frameworks Desirable to augment ML model predictions with expert inputs Popular examples – Healthcare models Content moderation tasks Useful for improving accuracy, incorporating additional information, and auditing models. Child maltreatment hotline screening
  • 151. Errors and biases in human-in-the-loop frameworks ML tasks often suffer from group-specific bias, induced due to misrepresentative data or models. Human-in-the-loop frameworks can reflect biases or inaccuracies of the human experts. Concerns include: • Racial bias in human-in-the-loop framework for recidivism risk assessment (Green, Chen – FAccT 2019) • Ethical concerns regarding audits of commercial facial processing technologies (Raji et al. – AIES 2020) • Automation bias in time critical decision support systems (Cummings – ISTC 2004) Can we design human-in-the-loop frameworks that take into account the expertise and biases of the human experts?
  • 152. Model 𝑋 − non-protected attributes; 𝑌 − class label; 𝑍 − protected attribute/group membership Number of experts available = 𝑚 − 1 Input −𝑋 Classifier 𝐹: 𝑋 → 𝑌 Deferrer 𝐷: 𝑋 → 0,1 𝑚 . . If 𝐷1 = 1 If 𝐷2 = 1 If 𝐷3 = 1 If 𝐷4 = 1 If 𝐷𝑚−1 = 1 If 𝐷𝑚 = 1 Deferrer 𝐷 choose a committee of experts. The majority decision of committee is the final prediction Final output of selected committee Experts might have access to additional information, including group membership 𝑍. There might be a cost/penalty associated with each expert review.
  • 153. Fairness in human-in-the-loop settings • Joint learning framework to learn a classifier and a deferral system for multiple experts simultaneously • Synthetic and real-world experiments on the efficacy of our method V. Keswani, M. Lease, K. Kenthapadi, Towards Unbiased and Accurate Deferral to Multiple Experts, AIES 2021.
  • 154. Minimax Group Fairness: Algorithms and Experiments • Equality of error rates: an intuitive & well-studied group fairness notion • May require artificially inflating error on easier-to-predict groups  • Undesirable when most/all of the targeted population is disadvantaged • Goal: minimize maximum group error [Martinez et al, 2020] • “Ensure that the worst-off group is as well-off as possible” • Our work: algorithms based on a zero-sum game between a Learner and a Regulator • Theoretical results and experimental evaluation E. Diana, W. Gill, M. Kearns, K. Kenthapadi, A. Roth, Minimax Group Fairness: Algorithms and Experiments, AIES 2021.
  • 155. Robust Interpretability of Neural Text Classifiers • Feature attribution methods used for understanding model predictions • How robust are feature attributions for neural text classifiers? • Are they identical under different random initializations of the same model? • Do they differ between a model with trained parameters and a model with random parameters? • Common feature attribution methods fail both tests! M. B. Zafar, M. Donini, D. Slack, C. Archambeau, S. Das, K. Kenthapadi, On the Lack of Robust Interpretability of Neural Text Classifiers, Findings in ACL 2021.
  • 156. Privacy Research @ Amazon - Sampler
  • 157. Differentially Private Query Release • Problem: answering marginal queries privately with high accuracy • Marginal queries are a special class of linear queries that count slices of a dataset. • “How many authors have visited Charlotte, graduated in the last two year and work in the Bay Area?” – A 3-way marginal query on the below dataset. • Privacy: Marginals computed against our dataset should protect against inferences on an individual’s membership (using Differential Privacy)
  • 158. Differentially Private Query Release • Projection Mechanism: Evaluate noisy answers to all queries in a query class Q and find the synthetic dataset (D′) in the space of feasible datasets that minimizes error with respect to some norm. Q is a class of queries. Nikolov, Talwar, Zhang, The geometry of differential privacy: the sparse and approximate cases, STOC 2013
  • 159. Differentially Private Query Release: Key Ideas 1. Relax the data domain: one-hot encode the non-continuous data and expand the domain to real numbers. Extend the differentiable queries to the new domain. 2. Adaptively select queries: repeatedly choose the k worst performing queries privately and optimize D′ to answer those well. S. Aydore, W. Brown, M. Kearns, K. Kenthapadi, L. Melis, A. Roth, A. Siva, Differentially Private Query Release Through Adaptive Projection, ICML 2021.
  • 160. Responsible AI Case Studies at Google Ben Packer Google AI
  • 162. Google Assistant Key Points: • Think about user harms How does your product make people feel • Adversarial ("stress") testing for all Google Assistant launches • People might say racist, sexist, homophobic stuff • Diverse testers • Think about expanding who your users could and should be • Consider the diversity of your users
  • 164. Google Camera Key points: • Check for unconscious bias • Comprehensive testing: "make sure this works for everybody"
  • 166. This is a “Shirley Card” Named after a Kodak studio model named Shirley Page, they were the primary method for calibrating color when processing film. SKIN TONE IN PHOTOGRAPHY SOURCES Color film was built for white people. Here's what it did to dark skin. (Vox) How Kodak's Shirley Cards Set Photography's Skin-Tone Standard, NPR
  • 167. Until about 1990, virtually all Shirley Cards featured Caucasian women. SKIN TONE IN PHOTOGRAPHY SOURCES Color film was built for white people. Here's what it did to dark skin. (Vox) Colour Balance, Image Technologies, and Cognitive Equity, Roth How Photography Was Optimized for White Skin Color (Priceonomics)
  • 168. As a result, photos featuring people with light skin looked fairly accurate. SKIN TONE IN PHOTOGRAPHY SOURCES Color film was built for white people. Here's what it did to dark skin. (Vox) Colour Balance, Image Technologies, and Cognitive Equity, Roth How Photography Was Optimized for White Skin Color (Priceonomics) Film Kodachrome Year 1970 Credit Darren Davis, Flickr
  • 169. Photos featuring people with darker skin, not so much... SKIN TONE IN PHOTOGRAPHY SOURCES Color film was built for white people. Here's what it did to dark skin. (Vox) Colour Balance, Image Technologies, and Cognitive Equity, Roth How Photography Was Optimized for White Skin Color (Priceonomics) Film Kodachrome Year 1958 Credit Peter Roome, Flickr
  • 171. Google Clips "We created controlled datasets by sampling subjects from different genders and skin tones in a balanced manner, while keeping variables like content type, duration, and environmental conditions constant. We then used this dataset to test that our algorithms had similar performance when applied to different groups." https://ai.googleblog.com/2018/05/automat ic-photography-with-google-clips.html
  • 172. Geena Davis Inclusion Quotient [with Geena Davis Institute on Gender in Media]
  • 174. Adversarial Testing for Smart Compose in Gmail
  • 177. Adversarial Testing for Smart Compose in Gmail
  • 181. 1. Detect Gender-Neutral Queries Train a text classifier to detect when a Turkish query is gender-neutral. • trained on thousands of human-rated Turkish examples
  • 182. 2. Generate Gender-Specific Translations • Training: Modify training data to add an additional input token specifying the required gender: • (<2MALE> O bir doktor, He is a doctor) • (<2FEMALE> O bir doktor, She is a doctor) • Deployment: If step (1) predicted query is gender-neutral, add male and female tokens to query • O bir doktor -> {<2MALE> O bir doktor, <2FEMALE> O bir doktor}
  • 183. 3. Check for Accuracy Verify: 1. If the requested feminine translation is feminine. 2. If the requested masculine translation is masculine. 3. If the feminine and masculine translations are exactly equivalent with the exception of gender-related changes.
  • 184. Result: Reduced Gender Bias in Translate
  • 186. Good ML Practices Go a Long Way Lots of low hanging fruit in terms of improving fairness simply by using machine learning best practices • Representative data • Introspection tools • Visualization tools • Testing 01 Fairness improvements often lead to overall improvements • It’s a common misconception that it’s always a tradeoff 02
  • 187. Breadth and Depth Required Looking End-to-End is critical • Need to be aware of bias and potential problems at every stage of product and ML pipelines (from design, data gathering, … to deployment and monitoring) 01 Details Matter • Slight changes in features or labeler criteria can change the outcome • Must have experts who understand the effects of decisions • Many details are not technical such as how labelers are hired 02
  • 188. Process Best Practices Identify product goals Get the right people in the room Identify stakeholders Select a fairness approach Analyze and evaluate your system Mitigate issues Monitor Continuously and Escalation Plans Auditing and Transparency Policy Technology
  • 189. Beyond Accuracy Performance and Cost Fairness and Bias Transparency and Explainability Privacy Security Safety Robustness
  • 191. Fairness in ML Application specific challenges Conversational AI systems: Unique bias/fairness/ethics considerations E.g., Hate speech, Complex failure modes Beyond protected categories, e.g., accent, dialect Entire ecosystem (e.g., including apps such as Alexa skills) Two-sided markets: e.g., fairness to buyers and to sellers, or to content consumers and producers Fairness in advertising (externalities) Tools for ensuring fairness (measuring & mitigating bias) in AI lifecycle Pre-processing (representative datasets; modifying features/labels) ML model training with fairness constraints Post-processing Experimentation & Post-deployment
  • 192. Key Open Problems in Applied Fairness What if you don’t have the sensitive attributes? When should you use what approach? For example, Equal treatment vs equal outcome? How to identify harms? Process for framing AI problems: Will the chosen metrics lead to desired results? How to tell if data generation and collection method is appropriate for a task? (e.g., causal structure analysis?) Processes for mitigating harms and misbehaviors quickly
  • 193. Explainability in ML Actionable explanations Balance between explanations & model secrecy Robustness of explanations to failure modes (Interaction between ML components) Application-specific challenges Conversational AI systems: contextual explanations Gradation of explanations Tools for explanations across AI lifecycle Pre & post-deployment for ML models Model developer vs. End user focused
  • 194. Privacy in ML Privacy for highly sensitive data: model training & analytics using secure enclaves, homomorphic encryption, federated learning / on- device learning, or a hybrid Privacy-preserving model training, robust against adversarial membership inference attacks (Dynamic settings + Complex data / model pipelines) Privacy-preserving mechanisms for data marketplaces
  • 195. Reflections “Fairness, Explainability, and Privacy by Design” when building AI products Collaboration/consensus across key stakeholders NYT / WSJ / ProPublica test :)
  • 196. Related Tutorials / Resources • ACM Conference on Fairness, Accountability, and Transparency (ACM FAccT) • AAAI/ACM Conference on Artificial Intelligence, Ethics, and Society (AIES) • Sara Hajian, Francesco Bonchi, and Carlos Castillo, Algorithmic bias: From discrimination discovery to fairness-aware data mining, KDD Tutorial, 2016. • Solon Barocas and Moritz Hardt, Fairness in machine learning, NeurIPS Tutorial, 2017. • Kate Crawford, The Trouble with Bias, NeurIPS Keynote, 2017. • Arvind Narayanan, 21 fairness definitions and their politics, FAccT Tutorial, 2018. • Sam Corbett-Davies and Sharad Goel, Defining and Designing Fair Algorithms, Tutorials at EC 2018 and ICML 2018. • Ben Hutchinson and Margaret Mitchell, Translation Tutorial: A History of Quantitative Fairness in Testing, FAccT Tutorial, 2019. • Henriette Cramer, Kenneth Holstein, Jennifer Wortman Vaughan, Hal Daumé III, Miroslav Dudík, Hanna Wallach, Sravana Reddy, and Jean Garcia-Gathright, Translation Tutorial: Challenges of incorporating algorithmic fairness into industry practice, FAccT Tutorial, 2019.
  • 197. Related Tutorials / Resources • Sarah Bird, Ben Hutchinson, Krishnaram Kenthapadi, Emre Kiciman, Margaret Mitchell, Fairness-Aware Machine Learning: Practical Challenges and Lessons Learned, Tutorials at WSDM 2019, WWW 2019, KDD 2019. • Krishna Gade, Sahin Cem Geyik, Krishnaram Kenthapadi, Varun Mithal, Ankur Taly, Explainable AI in Industry, Tutorials at KDD 2019, FAccT 2020, WWW 2020. • Himabindu Lakkaraju, Julius Adebayo, Sameer Singh, Explaining Machine Learning Predictions: State-of-the-art, Challenges, and Opportunities, NeurIPS 2020 Tutorial. • Kamalika Chaudhuri, Anand D. Sarwate, Differentially Private Machine Learning: Theory, Algorithms, and Applications, NeurIPS 2017 Tutorial. • Krishnaram Kenthapadi, Ilya Mironov, Abhradeep Guha Thakurta, Privacy-preserving Data Mining in Industry, Tutorials at KDD 2018, WSDM 2019, WWW 2019.