Abstract image of a woman sitting on books and studying on a laptop

Keynote: Bias in Search and Recommender Systems

Catalyst, profile picture
Catalyst

This document summarizes a keynote presentation about biases in search and recommender systems. The presentation discusses different types of biases that can occur at various stages, from data biases to algorithmic and interaction biases. It provides examples of how biases can manifest, such as popularity bias that disadvantages long-tail content in recommender systems. The presentation also discusses challenges in evaluating systems due to biases in test data and proposes methods for mitigating biases, such as stratified sampling and debiasing algorithms. The overarching message is that awareness of biases is important for building more transparent, accountable and fair systems.

Marketing
1 of 55
Downloaded 65 times
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
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
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost #TechSEOBoost | @CatalystSEM THANK YOU TO THIS YEAR’S SPONSORS Keynote: Bias in Search and Recommender Systems Ricardo Baeza-Yates, NTENT
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Ricardo Baeza-Yates CTO, NTENT Biases in Search & Recommender Systems
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost NTENT ntent.com Marketing Engineering Operations International Applied Research ntent.com
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Prologue
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost A Bit of History Data Volume Complexity IR DB Two different points of view for data
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Data Understanding Data Query Unstructured Structured Explicit Information Retrieval (Relational) Databases Implicit Recommender Systems Unknown Data Mining
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost What is Bias? • Statistical: significant systematic deviation from a prior (unknown) distribution; • Cultural: interpretations and judgments phenomena acquired through our life; • Cognitive: systematic pattern of deviation from norm or rationality in judgment;
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost So (Observational) Human Data has Bias • Gender • Racial • Sexual • Age • Religious • Social • Linguistic • Geographic • Political • Educational • Economic • Technological ▪ Gathering process ▪ Sampling process ▪ Validity (e.g. temporal) ▪ Completeness ▪ Noise, spam Many people extrapolate results of a sample to the whole population (e.g., social media analysis) In addition there is bias when measuring bias as well as bias towards measuring it! Attempt of an unbiased (personal) view on bias in Search & RS Cultural Biases Statistical Biases Cognitive Biases Self-selection
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Impact in Search and Recommender Systems • Most web systems are optimized by using implicit user feedback • However, user data is partly biased to the choices that these systems make • Clicks can only be done on things that are shown to us • As those systems are usually based in ML, they learn to reinforce their own biases, yielding self-fulfilled prophecies and/or sub-optimal solutions • For example, personalization and filter bubbles for users • but also echo chambers for (recommender) systems • Moreover, sometimes these systems compete among themselves, learning also biases of other systems rather than real user behavior • Even more, an improvement in one system might be just a degradation in another system that uses a different (even inversely correlated) optimization function • For example, user experience vs. monetization
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost A Non-Technical Question Algorithm Biased Data Neutral? Fair? Same Bias Garbage In Garbage Out
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost What is being fair?
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost A Non-Technical Question Algorithm Biased Data Neutral? Fair? Same Bias Not Always! Debias the input Tune the algorithm Debias the output Bias awareness!
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost January 2017 ACM US Statement on Algorithm Transparency and Accountability 1. Awareness 2. Access and redress 3. Accountability 4. Explanation 5. Data Provenance 6. Auditability 7. Validation and Testing Systems do not need to be perfect, they just need to be better than us
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Biases Everywhere!
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Data bias Biases on Search & RS: Web Case Study Web Spam
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost [Baeza-Yates, Castillo & López, Cybermetrics, 2005] Number of linked domains Exports(thousandsofUS$) Economic Bias in Links
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost 17 [Baeza-Yates & Castillo, WWW2006] Economic Bias in Links
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost 18 Minimal effortShameCultural Bias in Websites [Baeza-Yates, Castillo, Efthimiadis, TOIT 2007]
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Language Bias
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost [Bolukbasi at al, NIPS 2016] • Word embedding’s in w2vNEWS Yes, about 60 to 70% at work although at college is the inverse Gender Bias in Content Most journalists are men?
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost [E. Graells-Garrido et al,. ACM Hypertext’15] Systemic bias? Equal opportunity? Gender Bias in Content Wikipedia Partial information
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Data bias Activity bias Bias on Usage Actions People We are all in the long tail! [Goel et al., WSDM 2010]
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Popularity Bias in Recommender Systems Items Users Popular items Rest of items (long tail) • Take care to recommended items that are not too popular • Metrics • Novelty enhancement • Problem solved! …really? 𝑛𝑜𝑣 𝑖 = 1 − # ratings of 𝑖 # users Items #interactions More novel Less novel 𝑎 𝑏 [Vargas & Castells, RecSys 2011]
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost A Self-fulfilling Prophecy? Popular items (short head) Rest of items (long tail) Observed user-item interaction Unobserved preference Items Users Ratings are missing not at random (MNAR)
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Test data (relevant items) Training data Unobserved preference Items Users Popular items (short head) Rest of items (long tail) avg P@𝑘 ∼ + 𝑘 Judgments are missing not at random (MNAR) Worse yet: user-system reinforcement loop (more later) A Self-fulfilling Prophecy? [Marlin et al., RecSys 2010] [Steck, RecSys 2010, 2011] [Fleder & Hossanagar, Management Sciences 2009]
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost 1.E-06 1.E-05 1.E-04 1.E-03 0 0.2 0.4 0.6 0.8 1 A Problem for IR Evaluation Methodology! 30 TREC collections Items / documents #ratings/judgments (infraction) [Bellogín, Castells & Cantador, IRJ 2017] To how many queries is a document relevant? 25% queries can be answered with less than 1% of the URLs! [Baeza-Yates, Boldi, Chierichetti, WWW 2015]
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Get Rid of the Popularity Bias! • In the rating split [Bellogín, Castells & Cantador, IRJ 2017] • In the metrics • Stratified recall [Steck, RecSys 2011] • Importance propensity scoring [Yang et al., RecSys 2018] • In the algorithms [Steck, RecSys 2011] [Lobato et al., ICML 2014] [Jannach et al., UMUAI 2015] [Cañamares & Castells, SIGIR 2018, best paper award] Test data (relevant items) Training data Unobserved preference Items Items #ratings Flat test Popularity strata #ratings Time Temporal split
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Activity Bias also Affects Content [Baeza-Yates & Saez-Trumper, ACM Hypertext 2015] Most users are passive (i.e., more than 90% are lurkers) Then, which percentage of active users produce 50% of the content?
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost October 2015
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost [Baeza-Yates & Saez-Trumper, ACM Hypertext 2015] Which percentage of active users produce 50% of the content? Wisdom of crowds is a partial illusion Activity Bias: The Wisdom of a Few
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Content Never Seen [Baeza-Yates & Saez-Trumper, ACM Hypertext 2015] The Digital Desert
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Data bias Activity bias Sampling (size) bias Algorithmic bias Search or Recommender System Bias on the Web
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost [E. Graells-Garrido & M. Lalmas, ACM Hypertext’14] Geographical Bias in Recommender Systems
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost • If we want to estimate the frequency of queries that appear with probability at least p with a certain relative error, ∊, we can use the standard binomial error formula which works well for p near ½ but not for p near 0 • Better is the Agresti-Coull technique (also called Take 2) which gives: where Z is the inverse of the standard normal distribution, 1 − 𝛼 is the confidence interval and • If p = 0.1, 1 − 𝛼 is 80% and ∊ is 10%, the standard formula gives n = 900, while with A-C we get n = 2342. [Brown, Cai & DasGupta, Statistical Science, 2001] [Baeza-Yates, SIGIR 2015, Industry track] Sample Size?
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost • Standard technique: • A good sample should cover well all the items distribution, but this does not work with very skewed distributions. 10 0 10 1 10 2 10 3 Rank 10 0 10 1 10 2 10 3 10 4 10 5 Frequency 50M 10M 100K 1K 10 0 10 1 10 2 10 3 10 4 10 5 10 6 10 7 Rank 10 0 10 1 10 2 10 3 10 4 10 5 Frequency 50M 10M 100K 1K [Zaragoza et al, CIKM 2010] Sampling Queries
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost 36 Stratified Sampling Example [Baeza-Yates, SIGIR 2015, Industry track]
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Data bias Activity bias Sampling bias Interaction bias (Self) selection bias Bias in the User Interaction Search or Recommender System
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Position bias Ranking bias Presentation or exposure bias Social bias Interaction bias Bias in the Interaction Amazon.com
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Position bias Presentation bias Social bias Interaction bias Ranking bias Click bias Scrolling bias Mouse movement bias Data and algorithmic bias Self-selection bias Dependencies: A Cascade of Biases!
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Ranking Bias in Web Search [Mediative Study, 2014]
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Ranking Bias: Click Bias in Web Search • Ranking & next page bias Navigational queries
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost CTR (log) 1 11 21 Rank Learning to Rank with bias [Joachims et al., WSDM 2017, best paper] + many other papers Fair rankings [Zehlike et al., CIKM 2017] Clicks as implicit positive user feedback Debiasing Search Clicks and Other Biases [Dupret & Piwowarski, SIGIR 2008] [Chapelle & Zhang, WWW 2009] [Dupret & Liao, WSDM 2010, best paper] Debias the input Tune the algorithm Debias the output
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Data bias Activity bias Sampling bias Interaction bias (Self) selection bias Second-order bias Vicious Cycle of Bias Search or Recommender System
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost [Baeza-Yates, Pereira & Ziviani, WWW 2008] Person Web content is redundant (> 20%) Query Ranking bias in new content Redundancy grows (35%) Search results New page Second Order Bias in Web Content [Fortunato, Flammini, Menczer & Vespignani. PNAS 2006]
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Data bias Activity bias Sampling bias Interaction bias (Self) selection bias Vicious Cycle of Bias Search or Recommender System Feedback loop bias
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Bias due to Personalization • The effect of self-selection bias • Avoid the rich get richer and poor get poorer syndrome • Avoid the echo chamber by empowering the tail Cold start problem solution: Explore & Exploit Partial solutions: • Diversity • Novelty • Serendipity • My dark side Wikipedia [Eli Pariser, Penguin 2011]
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Users’ Eco Chambers in Recommender Systems ▪ Filter bubbles ▪ Degenerate feedback loops (e.g., YouTube autoplay) [Jiang et al., AAAI 2019]
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Eco Chamber of the Recommender System • Short-term greedy optimization, partial knowledge of the world • Long-term revenue optimization is not achieved • Views from new users should balance the exploration for new items • Disparate impact: unfair ecommerce/information markets [Baeza-Yates & Ribaudo, to appear]
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Epilogue
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Recap Bias Type Statistical Cultural Cognitive Algorithmic  ? ? Presentation  Position    Data   Sampling    Activity  Self-selection   Interaction   Social   Second order   
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost [Silberzahn et al., COS, Univ. of Virginia, 2015] Professional Bias? ➔ 61 analysts, 29 teams: 20 yes and 9 no
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost If Systems Reflects Its Designers: What we can/should do? ▪ Data ▪ Analyze for known and unknown biases, debias when possible/needed ▪ Recollect more data for difficult/sparse regions of the problem ▪ Delete attributes associated directly/indirectly with harmful bias ▪ Interaction ▪ Make sure that the user is aware of the biases all the time ▪ Give more control to the user ▪ Design and Implementation • Let experts/colleagues/users contest every step of the process ▪ Evaluation • Do not fool yourself!
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Final Take-Home Message ▪ Systems are a mirror of us, the good, the bad and the ugly ▪ The Web amplifies everything, but always leaves traces ▪ We need to be aware of our own biases! ▪ We must be aware of the biases and contrarrest them to stop the vicious bias cycle ▪ Plenty of open (research) problems! Big Data of People is huge….. ….. but it is tiny compared to the future Big Data of the Internet of Things (IoT) No activity bias!
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Thank You! Any Questions? – rbaeza@acm.org | http://www.baeza.cl/ | http://fairness-measures.org Biased Questions? ASIST 2012 Book of the Year Award (Biased Ad)
Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Thanks for Viewing the Slideshare! – Watch the Recording: https://youtube.com/session-example Or Contact us today to discover how Catalyst can deliver unparalleled SEO results for your business. https://www.catalystdigital.com/

More Related Content

PDF
Tutorial on Bias in Rec Sys @ UMAP2020
Mirko Marras
 
PPTX
Fair Recommender Systems
Sharmistha Chatterjee
 
PDF
Tutorial on Advances in Bias-aware Recommendation on the Web @ WSDM 2021
Mirko Marras
 
PDF
Fairness and Popularity Bias in Recommender Systems: an Empirical Evaluation
Cataldo Musto
 
PDF
An introduction to Recommender Systems
David Zibriczky
 
PDF
Learning to Rank for Recommender Systems - ACM RecSys 2013 tutorial
Alexandros Karatzoglou
 
PDF
Deep Learning for Recommender Systems
Justin Basilico
 
PDF
Overview of recommender system
Stanley Wang
 
Tutorial on Bias in Rec Sys @ UMAP2020
Mirko Marras
 
Fair Recommender Systems
Sharmistha Chatterjee
 
Tutorial on Advances in Bias-aware Recommendation on the Web @ WSDM 2021
Mirko Marras
 
Fairness and Popularity Bias in Recommender Systems: an Empirical Evaluation
Cataldo Musto
 
An introduction to Recommender Systems
David Zibriczky
 
Learning to Rank for Recommender Systems - ACM RecSys 2013 tutorial
Alexandros Karatzoglou
 
Deep Learning for Recommender Systems
Justin Basilico
 
Overview of recommender system
Stanley Wang
 

What's hot (20)

PPTX
Recommender system
Nilotpal Pramanik
 
PPTX
Recommender system introduction
Liang Xiang
 
PDF
Tutorial: Context In Recommender Systems
YONG ZHENG
 
PDF
Recommender Systems
Carlos Castillo (ChaTo)
 
PPTX
Content based filtering
Bendito Freitas Ribeiro
 
PDF
Past, present, and future of Recommender Systems: an industry perspective
Xavier Amatriain
 
PDF
Graph Neural Networks for Recommendations
WQ Fan
 
PDF
Active Learning in Collaborative Filtering Recommender Systems : a Survey
University of Bergen
 
PDF
Recommender Systems
Francesco Casalegno
 
PPTX
Recommender Systems
Lior Rokach
 
PDF
Learning to rank
Bruce Kuo
 
PPT
Diversity and novelty for recommendation system
Zhenv5
 
PPTX
Tag based recommender system
Karen Li
 
PPTX
Tutorial on sequence aware recommender systems - UMAP 2018
Paolo Cremonesi
 
PPTX
Recommendation Systems
Robin Reni
 
PDF
Context-aware Recommendation: A Quick View
YONG ZHENG
 
PPTX
Netflix talk at ML Platform meetup Sep 2019
Faisal Siddiqi
 
PDF
A Multi-Armed Bandit Framework For Recommendations at Netflix
Jaya Kawale
 
PDF
Tutorial on Sequence Aware Recommender Systems - ACM RecSys 2018
Massimo Quadrana
 
PDF
Recommendation System Explained
Crossing Minds
 
Recommender system
Nilotpal Pramanik
 
Recommender system introduction
Liang Xiang
 
Tutorial: Context In Recommender Systems
YONG ZHENG
 
Recommender Systems
Carlos Castillo (ChaTo)
 
Content based filtering
Bendito Freitas Ribeiro
 
Past, present, and future of Recommender Systems: an industry perspective
Xavier Amatriain
 
Graph Neural Networks for Recommendations
WQ Fan
 
Active Learning in Collaborative Filtering Recommender Systems : a Survey
University of Bergen
 
Recommender Systems
Francesco Casalegno
 
Recommender Systems
Lior Rokach
 
Learning to rank
Bruce Kuo
 
Diversity and novelty for recommendation system
Zhenv5
 
Tag based recommender system
Karen Li
 
Tutorial on sequence aware recommender systems - UMAP 2018
Paolo Cremonesi
 
Recommendation Systems
Robin Reni
 
Context-aware Recommendation: A Quick View
YONG ZHENG
 
Netflix talk at ML Platform meetup Sep 2019
Faisal Siddiqi
 
A Multi-Armed Bandit Framework For Recommendations at Netflix
Jaya Kawale
 
Tutorial on Sequence Aware Recommender Systems - ACM RecSys 2018
Massimo Quadrana
 
Recommendation System Explained
Crossing Minds
 
Ad

Similar to Keynote: Bias in Search and Recommender Systems (20)

PDF
Ux day2018 ricardo baeza yayes search-biases-semantics
Multiplica
 
PDF
Data and Algorithmic Bias in the Web
WebVisions
 
PDF
In a World of Biased Search Engines
Dirk Lewandowski
 
PPTX
Fairness in Search & RecSys 네이버 검색 콜로키움 김진영
Jin Young Kim
 
PPT
Recommender Systems Tutorial (Part 1) -- Introduction
Bee-Chung Chen
 
PPTX
How Humans & Machines Can Improve Site Search Results - Search Y: Paris
JP Sherman
 
PPTX
Algorithms of Online Platforms and Networks
Ansgar Koene
 
PDF
Bias in recommendation: avoid it or embrace it?
Pablo Castells
 
PPTX
Deconstructing Popularity Bias in Recommender Systems_ Origins, Impacts, and ...
Amit Jaspal
 
PPTX
Not fair! testing AI bias and organizational values
Peter Varhol
 
PPTX
Measuring effectiveness of machine learning systems
Amit Sharma
 
PDF
Challenges in Evaluating Exploration Effectiveness in Recommender Systems
Inbar Naor
 
PPTX
Not fair! testing ai bias and organizational values
Peter Varhol
 
PPTX
Immersive Recommendation Workshop, NYC Media Lab'17
Longqi Yang
 
PDF
Measuring Relevance in the Negative Space
Trey Grainger
 
PDF
AI day poster
Joyce Chou
 
PDF
Bias in AI-systems: A multi-step approach
Eirini Ntoutsi
 
PDF
AI ETHICS AND BIASES (For the AI BIAS).pdf
AnwaarAlam2
 
PDF
Data science toolkit for product managers
ProductFolks
 
PDF
Data Science Toolkit for Product Managers
Mahmoud Jalajel
 
Ux day2018 ricardo baeza yayes search-biases-semantics
Multiplica
 
Data and Algorithmic Bias in the Web
WebVisions
 
In a World of Biased Search Engines
Dirk Lewandowski
 
Fairness in Search & RecSys 네이버 검색 콜로키움 김진영
Jin Young Kim
 
Recommender Systems Tutorial (Part 1) -- Introduction
Bee-Chung Chen
 
How Humans & Machines Can Improve Site Search Results - Search Y: Paris
JP Sherman
 
Algorithms of Online Platforms and Networks
Ansgar Koene
 
Bias in recommendation: avoid it or embrace it?
Pablo Castells
 
Deconstructing Popularity Bias in Recommender Systems_ Origins, Impacts, and ...
Amit Jaspal
 
Not fair! testing AI bias and organizational values
Peter Varhol
 
Measuring effectiveness of machine learning systems
Amit Sharma
 
Challenges in Evaluating Exploration Effectiveness in Recommender Systems
Inbar Naor
 
Not fair! testing ai bias and organizational values
Peter Varhol
 
Immersive Recommendation Workshop, NYC Media Lab'17
Longqi Yang
 
Measuring Relevance in the Negative Space
Trey Grainger
 
AI day poster
Joyce Chou
 
Bias in AI-systems: A multi-step approach
Eirini Ntoutsi
 
AI ETHICS AND BIASES (For the AI BIAS).pdf
AnwaarAlam2
 
Data science toolkit for product managers
ProductFolks
 
Data Science Toolkit for Product Managers
Mahmoud Jalajel
 
Ad

More from Catalyst (20)

PDF
Closing the Gap: Adopting Omnichannel Strategies for Stronger Brand-Consumer ...
Catalyst
 
PDF
TechSEO Boost 2021 - Cultivating a Product Mindset for Success
Catalyst
 
PDF
TechSEO Boost 2021 - SEO Experimentation
Catalyst
 
PDF
TechSEO Boost 2021 - Rendering Strategies: Measuring the Devil’s Details in C...
Catalyst
 
PDF
TechSEO Boost 2021 - The Future Is The Past: Tagging And Tracking Through The...
Catalyst
 
PDF
10 Trends Changing Programmatic
Catalyst
 
PDF
New Commerce Conference: Charting a Course to Success with Your Retail Media ...
Catalyst
 
PDF
The New Commerce Conference: The Omni-channel Imperative
Catalyst
 
PDF
New Commerce Commerce: All Things Instacart
Catalyst
 
PDF
The Power of SEO: Protect Your Bottom Line & Future Proof Your Brand
Catalyst
 
PDF
The Era of Omni-Commerce: New Insights for Dominating the Digital Shelf and B...
Catalyst
 
PDF
Reignite Your Business with Performance Marketing: 4 Ways to Fuel Your Reopening
Catalyst
 
PDF
Reignite Your Business with Performance Marketing: 4 Ways to Dial-Up Brand In...
Catalyst
 
PDF
Evolve Your Social Commerce Strategy: Thinking Beyond Facebook
Catalyst
 
PDF
B2B SEO: Increase Traffic & Leads in 2020
Catalyst
 
PDF
TechSEO Boost 2019: Research Competition
Catalyst
 
PDF
NLP Powered Outreach Link Building
Catalyst
 
PDF
Generating Qualitative Content with GPT-2 in All Languages
Catalyst
 
PDF
Automate, Create Tools, & Test Ideas Quickly with Google Apps Script
Catalyst
 
PDF
NLP for SEO
Catalyst
 
Closing the Gap: Adopting Omnichannel Strategies for Stronger Brand-Consumer ...
Catalyst
 
TechSEO Boost 2021 - Cultivating a Product Mindset for Success
Catalyst
 
TechSEO Boost 2021 - SEO Experimentation
Catalyst
 
TechSEO Boost 2021 - Rendering Strategies: Measuring the Devil’s Details in C...
Catalyst
 
TechSEO Boost 2021 - The Future Is The Past: Tagging And Tracking Through The...
Catalyst
 
10 Trends Changing Programmatic
Catalyst
 
New Commerce Conference: Charting a Course to Success with Your Retail Media ...
Catalyst
 
The New Commerce Conference: The Omni-channel Imperative
Catalyst
 
New Commerce Commerce: All Things Instacart
Catalyst
 
The Power of SEO: Protect Your Bottom Line & Future Proof Your Brand
Catalyst
 
The Era of Omni-Commerce: New Insights for Dominating the Digital Shelf and B...
Catalyst
 
Reignite Your Business with Performance Marketing: 4 Ways to Fuel Your Reopening
Catalyst
 
Reignite Your Business with Performance Marketing: 4 Ways to Dial-Up Brand In...
Catalyst
 
Evolve Your Social Commerce Strategy: Thinking Beyond Facebook
Catalyst
 
B2B SEO: Increase Traffic & Leads in 2020
Catalyst
 
TechSEO Boost 2019: Research Competition
Catalyst
 
NLP Powered Outreach Link Building
Catalyst
 
Generating Qualitative Content with GPT-2 in All Languages
Catalyst
 
Automate, Create Tools, & Test Ideas Quickly with Google Apps Script
Catalyst
 
NLP for SEO
Catalyst
 

Recently uploaded (20)

PPTX
Week-1-BM.pptx for student that need to study
glesybelgerolaga
 
PDF
Session 4 - Segmenting Business Market & Estimating Demand.pdf
mukulr530
 
PDF
OOH Platforms in Spain Data-Driven Strategies for Better Campaigns.pdf
adcities
 
PDF
SEO Is Alive: Real Data That Kills the Internet Hysteria - Sid Lal, Bruce Cla...
DigiMarCon - Digital Marketing, Media and Advertising Conferences & Exhibitions
 
PPTX
Quizon logo and tagline recognition.pptx
Class1013
 
PDF
The Internet of Agents - Alexander De Ridder, SmythOS
DigiMarCon - Digital Marketing, Media and Advertising Conferences & Exhibitions
 
PPTX
IMC Bimtech --------------------------.pptx
kashish261
 
PDF
EYP Creation Presentation Deck - Offerings
Arvind Mishra
 
PDF
Critical Digital Marketing Shifts Businesses Need to Act on Now - Marina Garb...
DigiMarCon - Digital Marketing, Media and Advertising Conferences & Exhibitions
 
PPTX
Best Artificial Intelligence Course in Jalandhar
PreetiSharma477899
 
PDF
BETRIMEX market penetration proposal - GROUP 3 - CANADA.pdf
mvttrang2201
 
PDF
Experiment Your Way to Page One: A Guide to SEO Testing
Ryan Jones
 
PDF
domain and Hosting by mayank adhikari ppt
manuadhikari36
 
DOCX
Creative Marketing Campaigns in Milton Keynes
lavanyawrites99
 
PDF
GEO vs SEO: Maximizing Engagement with LLM
valeria967173
 
PPT
david_sm13_ppt_01.ppt MARKETING MANAGEMENT
poojamehra52
 
PPTX
DOC-20241015-WA0008. (1).pptx hotel management
Laxmibadiger8
 
PPTX
The principles of Marketing Environment 2.pptx
GoharSaeed6
 
DOCX
How Can Faster Websites Enhance Customer Experience in Milton Keynes?
lavanyawrites99
 
PPTX
Best Machine & AI Company in India - Digital Navik
learningstudio62
 
Week-1-BM.pptx for student that need to study
glesybelgerolaga
 
Session 4 - Segmenting Business Market & Estimating Demand.pdf
mukulr530
 
OOH Platforms in Spain Data-Driven Strategies for Better Campaigns.pdf
adcities
 
SEO Is Alive: Real Data That Kills the Internet Hysteria - Sid Lal, Bruce Cla...
DigiMarCon - Digital Marketing, Media and Advertising Conferences & Exhibitions
 
Quizon logo and tagline recognition.pptx
Class1013
 
The Internet of Agents - Alexander De Ridder, SmythOS
DigiMarCon - Digital Marketing, Media and Advertising Conferences & Exhibitions
 
IMC Bimtech --------------------------.pptx
kashish261
 
EYP Creation Presentation Deck - Offerings
Arvind Mishra
 
Critical Digital Marketing Shifts Businesses Need to Act on Now - Marina Garb...
DigiMarCon - Digital Marketing, Media and Advertising Conferences & Exhibitions
 
Best Artificial Intelligence Course in Jalandhar
PreetiSharma477899
 
BETRIMEX market penetration proposal - GROUP 3 - CANADA.pdf
mvttrang2201
 
Experiment Your Way to Page One: A Guide to SEO Testing
Ryan Jones
 
domain and Hosting by mayank adhikari ppt
manuadhikari36
 
Creative Marketing Campaigns in Milton Keynes
lavanyawrites99
 
GEO vs SEO: Maximizing Engagement with LLM
valeria967173
 
david_sm13_ppt_01.ppt MARKETING MANAGEMENT
poojamehra52
 
DOC-20241015-WA0008. (1).pptx hotel management
Laxmibadiger8
 
The principles of Marketing Environment 2.pptx
GoharSaeed6
 
How Can Faster Websites Enhance Customer Experience in Milton Keynes?
lavanyawrites99
 
Best Machine & AI Company in India - Digital Navik
learningstudio62
 

Keynote: Bias in Search and Recommender Systems

  • 1. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost #TechSEOBoost | @CatalystSEM THANK YOU TO THIS YEAR’S SPONSORS Keynote: Bias in Search and Recommender Systems Ricardo Baeza-Yates, NTENT
  • 2. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Ricardo Baeza-Yates CTO, NTENT Biases in Search & Recommender Systems
  • 3. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost NTENT ntent.com Marketing Engineering Operations International Applied Research ntent.com
  • 4. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Prologue
  • 5. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost A Bit of History Data Volume Complexity IR DB Two different points of view for data
  • 6. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Data Understanding Data Query Unstructured Structured Explicit Information Retrieval (Relational) Databases Implicit Recommender Systems Unknown Data Mining
  • 7. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost What is Bias? • Statistical: significant systematic deviation from a prior (unknown) distribution; • Cultural: interpretations and judgments phenomena acquired through our life; • Cognitive: systematic pattern of deviation from norm or rationality in judgment;
  • 8. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost So (Observational) Human Data has Bias • Gender • Racial • Sexual • Age • Religious • Social • Linguistic • Geographic • Political • Educational • Economic • Technological ▪ Gathering process ▪ Sampling process ▪ Validity (e.g. temporal) ▪ Completeness ▪ Noise, spam Many people extrapolate results of a sample to the whole population (e.g., social media analysis) In addition there is bias when measuring bias as well as bias towards measuring it! Attempt of an unbiased (personal) view on bias in Search & RS Cultural Biases Statistical Biases Cognitive Biases Self-selection
  • 9. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Impact in Search and Recommender Systems • Most web systems are optimized by using implicit user feedback • However, user data is partly biased to the choices that these systems make • Clicks can only be done on things that are shown to us • As those systems are usually based in ML, they learn to reinforce their own biases, yielding self-fulfilled prophecies and/or sub-optimal solutions • For example, personalization and filter bubbles for users • but also echo chambers for (recommender) systems • Moreover, sometimes these systems compete among themselves, learning also biases of other systems rather than real user behavior • Even more, an improvement in one system might be just a degradation in another system that uses a different (even inversely correlated) optimization function • For example, user experience vs. monetization
  • 10. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost A Non-Technical Question Algorithm Biased Data Neutral? Fair? Same Bias Garbage In Garbage Out
  • 11. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost What is being fair?
  • 12. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost A Non-Technical Question Algorithm Biased Data Neutral? Fair? Same Bias Not Always! Debias the input Tune the algorithm Debias the output Bias awareness!
  • 13. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost January 2017 ACM US Statement on Algorithm Transparency and Accountability 1. Awareness 2. Access and redress 3. Accountability 4. Explanation 5. Data Provenance 6. Auditability 7. Validation and Testing Systems do not need to be perfect, they just need to be better than us
  • 14. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Biases Everywhere!
  • 15. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Data bias Biases on Search & RS: Web Case Study Web Spam
  • 16. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost [Baeza-Yates, Castillo & López, Cybermetrics, 2005] Number of linked domains Exports(thousandsofUS$) Economic Bias in Links
  • 17. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost 17 [Baeza-Yates & Castillo, WWW2006] Economic Bias in Links
  • 18. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost 18 Minimal effortShameCultural Bias in Websites [Baeza-Yates, Castillo, Efthimiadis, TOIT 2007]
  • 19. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Language Bias
  • 20. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost [Bolukbasi at al, NIPS 2016] • Word embedding’s in w2vNEWS Yes, about 60 to 70% at work although at college is the inverse Gender Bias in Content Most journalists are men?
  • 21. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost [E. Graells-Garrido et al,. ACM Hypertext’15] Systemic bias? Equal opportunity? Gender Bias in Content Wikipedia Partial information
  • 22. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Data bias Activity bias Bias on Usage Actions People We are all in the long tail! [Goel et al., WSDM 2010]
  • 23. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Popularity Bias in Recommender Systems Items Users Popular items Rest of items (long tail) • Take care to recommended items that are not too popular • Metrics • Novelty enhancement • Problem solved! …really? 𝑛𝑜𝑣 𝑖 = 1 − # ratings of 𝑖 # users Items #interactions More novel Less novel 𝑎 𝑏 [Vargas & Castells, RecSys 2011]
  • 24. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost A Self-fulfilling Prophecy? Popular items (short head) Rest of items (long tail) Observed user-item interaction Unobserved preference Items Users Ratings are missing not at random (MNAR)
  • 25. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Test data (relevant items) Training data Unobserved preference Items Users Popular items (short head) Rest of items (long tail) avg P@𝑘 ∼ + 𝑘 Judgments are missing not at random (MNAR) Worse yet: user-system reinforcement loop (more later) A Self-fulfilling Prophecy? [Marlin et al., RecSys 2010] [Steck, RecSys 2010, 2011] [Fleder & Hossanagar, Management Sciences 2009]
  • 26. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost 1.E-06 1.E-05 1.E-04 1.E-03 0 0.2 0.4 0.6 0.8 1 A Problem for IR Evaluation Methodology! 30 TREC collections Items / documents #ratings/judgments (infraction) [Bellogín, Castells & Cantador, IRJ 2017] To how many queries is a document relevant? 25% queries can be answered with less than 1% of the URLs! [Baeza-Yates, Boldi, Chierichetti, WWW 2015]
  • 27. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Get Rid of the Popularity Bias! • In the rating split [Bellogín, Castells & Cantador, IRJ 2017] • In the metrics • Stratified recall [Steck, RecSys 2011] • Importance propensity scoring [Yang et al., RecSys 2018] • In the algorithms [Steck, RecSys 2011] [Lobato et al., ICML 2014] [Jannach et al., UMUAI 2015] [Cañamares & Castells, SIGIR 2018, best paper award] Test data (relevant items) Training data Unobserved preference Items Items #ratings Flat test Popularity strata #ratings Time Temporal split
  • 28. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Activity Bias also Affects Content [Baeza-Yates & Saez-Trumper, ACM Hypertext 2015] Most users are passive (i.e., more than 90% are lurkers) Then, which percentage of active users produce 50% of the content?
  • 29. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost October 2015
  • 30. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost [Baeza-Yates & Saez-Trumper, ACM Hypertext 2015] Which percentage of active users produce 50% of the content? Wisdom of crowds is a partial illusion Activity Bias: The Wisdom of a Few
  • 31. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Content Never Seen [Baeza-Yates & Saez-Trumper, ACM Hypertext 2015] The Digital Desert
  • 32. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Data bias Activity bias Sampling (size) bias Algorithmic bias Search or Recommender System Bias on the Web
  • 33. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost [E. Graells-Garrido & M. Lalmas, ACM Hypertext’14] Geographical Bias in Recommender Systems
  • 34. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost • If we want to estimate the frequency of queries that appear with probability at least p with a certain relative error, ∊, we can use the standard binomial error formula which works well for p near ½ but not for p near 0 • Better is the Agresti-Coull technique (also called Take 2) which gives: where Z is the inverse of the standard normal distribution, 1 − 𝛼 is the confidence interval and • If p = 0.1, 1 − 𝛼 is 80% and ∊ is 10%, the standard formula gives n = 900, while with A-C we get n = 2342. [Brown, Cai & DasGupta, Statistical Science, 2001] [Baeza-Yates, SIGIR 2015, Industry track] Sample Size?
  • 35. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost • Standard technique: • A good sample should cover well all the items distribution, but this does not work with very skewed distributions. 10 0 10 1 10 2 10 3 Rank 10 0 10 1 10 2 10 3 10 4 10 5 Frequency 50M 10M 100K 1K 10 0 10 1 10 2 10 3 10 4 10 5 10 6 10 7 Rank 10 0 10 1 10 2 10 3 10 4 10 5 Frequency 50M 10M 100K 1K [Zaragoza et al, CIKM 2010] Sampling Queries
  • 36. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost 36 Stratified Sampling Example [Baeza-Yates, SIGIR 2015, Industry track]
  • 37. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Data bias Activity bias Sampling bias Interaction bias (Self) selection bias Bias in the User Interaction Search or Recommender System
  • 38. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Position bias Ranking bias Presentation or exposure bias Social bias Interaction bias Bias in the Interaction Amazon.com
  • 39. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Position bias Presentation bias Social bias Interaction bias Ranking bias Click bias Scrolling bias Mouse movement bias Data and algorithmic bias Self-selection bias Dependencies: A Cascade of Biases!
  • 40. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Ranking Bias in Web Search [Mediative Study, 2014]
  • 41. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Ranking Bias: Click Bias in Web Search • Ranking & next page bias Navigational queries
  • 42. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost CTR (log) 1 11 21 Rank Learning to Rank with bias [Joachims et al., WSDM 2017, best paper] + many other papers Fair rankings [Zehlike et al., CIKM 2017] Clicks as implicit positive user feedback Debiasing Search Clicks and Other Biases [Dupret & Piwowarski, SIGIR 2008] [Chapelle & Zhang, WWW 2009] [Dupret & Liao, WSDM 2010, best paper] Debias the input Tune the algorithm Debias the output
  • 43. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Data bias Activity bias Sampling bias Interaction bias (Self) selection bias Second-order bias Vicious Cycle of Bias Search or Recommender System
  • 44. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost [Baeza-Yates, Pereira & Ziviani, WWW 2008] Person Web content is redundant (> 20%) Query Ranking bias in new content Redundancy grows (35%) Search results New page Second Order Bias in Web Content [Fortunato, Flammini, Menczer & Vespignani. PNAS 2006]
  • 45. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Data bias Activity bias Sampling bias Interaction bias (Self) selection bias Vicious Cycle of Bias Search or Recommender System Feedback loop bias
  • 46. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Bias due to Personalization • The effect of self-selection bias • Avoid the rich get richer and poor get poorer syndrome • Avoid the echo chamber by empowering the tail Cold start problem solution: Explore & Exploit Partial solutions: • Diversity • Novelty • Serendipity • My dark side Wikipedia [Eli Pariser, Penguin 2011]
  • 47. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Users’ Eco Chambers in Recommender Systems ▪ Filter bubbles ▪ Degenerate feedback loops (e.g., YouTube autoplay) [Jiang et al., AAAI 2019]
  • 48. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Eco Chamber of the Recommender System • Short-term greedy optimization, partial knowledge of the world • Long-term revenue optimization is not achieved • Views from new users should balance the exploration for new items • Disparate impact: unfair ecommerce/information markets [Baeza-Yates & Ribaudo, to appear]
  • 49. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Epilogue
  • 50. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Recap Bias Type Statistical Cultural Cognitive Algorithmic  ? ? Presentation  Position    Data   Sampling    Activity  Self-selection   Interaction   Social   Second order   
  • 51. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost [Silberzahn et al., COS, Univ. of Virginia, 2015] Professional Bias? ➔ 61 analysts, 29 teams: 20 yes and 9 no
  • 52. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost If Systems Reflects Its Designers: What we can/should do? ▪ Data ▪ Analyze for known and unknown biases, debias when possible/needed ▪ Recollect more data for difficult/sparse regions of the problem ▪ Delete attributes associated directly/indirectly with harmful bias ▪ Interaction ▪ Make sure that the user is aware of the biases all the time ▪ Give more control to the user ▪ Design and Implementation • Let experts/colleagues/users contest every step of the process ▪ Evaluation • Do not fool yourself!
  • 53. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Final Take-Home Message ▪ Systems are a mirror of us, the good, the bad and the ugly ▪ The Web amplifies everything, but always leaves traces ▪ We need to be aware of our own biases! ▪ We must be aware of the biases and contrarrest them to stop the vicious bias cycle ▪ Plenty of open (research) problems! Big Data of People is huge….. ….. but it is tiny compared to the future Big Data of the Internet of Things (IoT) No activity bias!
  • 54. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Thank You! Any Questions? – [email protected] | http://www.baeza.cl/ | http://fairness-measures.org Biased Questions? ASIST 2012 Book of the Year Award (Biased Ad)
  • 55. Ricardo Baeza-Yates © | @polarbearby | #TechSEOBoost Thanks for Viewing the Slideshare! – Watch the Recording: https://youtube.com/session-example Or Contact us today to discover how Catalyst can deliver unparalleled SEO results for your business. https://www.catalystdigital.com/