HCI Research as Problem-Solving [CHI'16, presentation slides]
2 likes4,347 views
This essay presents a meta-scientific framework for understanding human-computer interaction (HCI) research as problem-solving, building on Larry Laudan's concepts of research problems and solutions. It categorizes HCI research into three types of problems—empirical, conceptual, and constructive—arguing for a unified approach that transcends traditional disciplinary boundaries. The authors contend that the progress of HCI can be assessed by its capacity to improve problem-solving in the context of human use of computers.
![A PAPER PRESENTED AT CHI 2016
Paper: http://users.comnet.aalto.fi/oulasvir/pubs/hci-research-as-problem-solving-chi2016.pdf
HCI Research as Problem-Solving
Antti Oulasvirta
Aalto University, Finland
Kasper Hornbæk
University of Copenhagen, Denmark
ABSTRACT
This essay contributes a meta-scientific account of human–
computer interaction (HCI) research as problem-solving. We
build on the philosophy of Larry Laudan, who develops
problem and solution as the foundational concepts of sci-
ence. We argue that most HCI research is about three main
types of problem: empirical, conceptual, and constructive.
We elaborate upon Laudan’s concept of problem-solving ca-
pacity as a universal criterion for determining the progress of
solutions (outcomes): Instead of asking whether research is
‘valid’ or follows the ‘right’ approach, it urges us to ask how
its solutions advance our capacity to solve important problems
in human use of computers. This offers a rich, generative, and
‘discipline-free’ view of HCI and resolves some existing de-
bates about what HCI is or should be. It may also help unify
efforts across nominally disparate traditions in empirical re-
search, theory, design, and engineering.
Author Keywords
Human–computer interaction; Problem-solving; Scientific
progress; Research problem; Larry Laudan
ACM Classification Keywords
H.5.m. Information interfaces and presentation (e.g., HCI):
Miscellaneous
INTRODUCTION
The spark for writing this essay comes from feelings of con-
fusion, and even embarrassment, arising in describing our
field to students and other researchers. What is human–com-
puter interaction (HCI) as a field? As numerous ideas and
disciplines contribute to HCI, its unique character is elusive.
Although HCI is in intellectual debt to many other fields, few
would agree that it reduces to them. It has its own subject of
enquiry, which is not part of the natural or social sciences. It
does not belong to engineering, computer science, or design
either. So what is it?
The essay has a grand ambition: to develop a conceptually
coherent account of the ‘95% of HCI research’. We know of
no other paper offering an attempt to address the field as a
whole. We are motivated first and foremost by the intellec-
tual enigma pertaining to what HCI is: There is no accepted
account that would tell how HCI’s numerous approaches
contribute to pursuit of shared objectives. In contrast, HCI
has been criticised for lack of ‘motor themes, mainstream
topics, and schools of thought’ [25] and for being fragmented
‘across topics, theories, methods, and people’ [38]. Conse-
quently, some have called for ‘a hard science’ [36], others
for ‘strong concepts’ [19] or an ‘inter-discipline’ [3]. These
are serious concerns with serious implications for the field.
Why bother with a meta-scientific paper at a technical con-
ference? Because the stakes are high. Philosophies of science
are at worst an impotent topic worthy of hallway conversa-
tions. But if the critics are right, our field is seriously crip-
pled, from the project level to the larger arenas of research
Realpolitik. Lacking a coherent view of what HCI is, and
what good research in HCI is, how can we communicate re-
sults to others, assess research, co-ordinate efforts, or com-
pete? In addition, as we show, philosophical views offer
thinking tools that can aid in generating ideas and generally
enhance the quality of research.
The contribution here lies in describing HCI as prob-
lem-solving. An overview is given in Figure 1. The view
originates from Larry Laudan’s philosophy of science [28].
Laudan describes scientific progress in terms of two founda-
tional concepts: research problem and solution. Laudan's
‘problem’ is not what we mean by the term in ordinary lan-
guage. It is defined via inabilities and absences occurring in
descriptions; knowledge; or, as often in HCI, constructive so-
lutions. For example, a research problem may involve lack
of understanding of how colour schemes on a web page af-
fect the aesthetic experience of its use. More generally, Lau-
dan’s research problem subsumes what we traditionally un-
derstand in HCI as a ‘design problem’ but also problems to
do with theory and empirical research.
Most of our argumentation builds on a concept put forth by
Laudan that links problems with solutions: problem-solving
capacity. For Laudan, a solution is something special, too. In
the above-mentioned case of aesthetic perception of web
pages, possible solutions range from descriptions of self-re-
ports to models of aesthetic impressions. These solutions
change the status of the inabilities and absences but in differ-
ent ways. Laudan qualifies this in terms of improvements to
problem-solving capacity. This is counter to some traditional
notions of progress [28, p. 14]:
In appraising the merits of theories, it is more important to
ask whether they constitute adequate solutions to signifi-
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. Copyrights for com-
ponents of this work owned by others than ACM must be honored. Ab-
stracting with credit is permitted. To copy otherwise, or republish, to post
on servers or to redistribute to lists, requires prior specific permission and/or
a fee. Request permissions from Permissions@acm.org.
CHI'16, May 07-12, 2016, San Jose, CA, USA
© 2016 ACM. ISBN 978-1-4503-3362-7/16/05…$15.00
DOI: http://dx.doi.org/10.1145/2858036.2858283
cant problems than it is to ask whether they are ‘true’, ‘cor-
roborated’, ‘well-confirmed’ or otherwise justifiable
within the framework of contemporary epistemology.
With this definition, the benefit of problem-solving is that it
allows covering a wider scope of research than previous ac-
counts, which have been restricted to certain disciplines, top-
ics, or approaches (e.g., research-through-design [53], inter-
action criticism [2], usability science [15], or interaction sci-
ence [21]). However, because Laudan developed his view
with natural and social sciences in mind, he missed design
and engineering contributions. Extending Laudan’s typology
to propose that research problems in HCI include not only
empirical and conceptual but also constructive problems, we
present the first typology developed to encompass most rec-
ognised research problems in HCI. It is now possible to de-
scribe research contributions regardless of the background
traditions, paradigms, or methods. The seemingly multi- or,
rather, hyper-disciplinary field is—in the end—about solv-
ing three types of problem. This reduces the number of di-
mensions dramatically when one is talking about HCI.
Having built the conceptual foundation, we return to answer
four fundamental questions: 1) What is HCI research, 2)
what is good HCI research, 3) are we doing a good job as a
field, and 4) could we do an even better job?
We aim to show through these discussions that Laudan's
problem-solving view is not just ‘solutionism’. It offers a
useful, timeless, and actionable non-disciplinary stance to
HCI. Instead of asking whether research subscribes to the
‘right’ approach, a system is ‘novel’, or a theory is ‘true’, one
asks how it advances our ability to solve important problems
relevant to human use of computers. Are we addressing the
right problems? Are we solving them well? The view helps
us contribute to some longstanding debates about HCI.
Moreover, we show that the view is generative. We provide
ideas on how to apply it as a thinking tool. Problem-solving
capacity can be analysed for individual papers or even whole
sub-topics and the field at large. It also works as a spring-
board for generating ideas to improve research agendas.
We conclude on a positive note by arguing that HCI is nei-
ther unscientific nor non-scientific (as some have claimed
[40]) or in deep crisis [25]. Such views do not recognise the
kinds of contributions being made. Instead, on many counts,
HCI has improved problem-solving capacity in human use of
computing remarkably and continues to do so. However, as
we show, these contributions tend to focus on empirical and
constructive problem types. In a contrast to calls for HCI to
be more scientific [21], interdisciplinary [3], hard [36], soft
[9], or rigorous [40], the systematic weakness of HCI is, in
fact, our inability to produce conceptual contributions (theo-
ries, methods, concepts, and principles) that link empirical
and constructive research.
THREE TYPES OF RESEARCH PROBLEM IN HCI
Our first point is that the key to understanding HCI as prob-
lem-solving is the recognition that its research efforts cluster
around a few recurring problem types. We effectively ‘col-
lapse’ the (apparent) multiplicity of research efforts under a
few problem types. This not only simplifies HCI but also
transcends some biasing presumptions arising from method-
ology, theory, or discipline. One can now see similarities and
differences between, say, an observational study of a novel
technology and a rigorous laboratory experiment, without
being bound by their traditions.
In this section, we 1) introduce Laudan’s notion of research
problem briefly, 2) extend his typology to cover engineering
and design contributions to HCI, and 3) argue that contribu-
tions in HCI can be classified via this typology.
Laudan originally distinguished only two types of research
problem—empirical and conceptual. These are defined in
terms of absence or inabilities to understand or achieve some
ends. As we argue below, the two types are applicable also
to HCI. However, to not let design ‘off the hook’, HCI should
cover engineering and design contributions. This aspect is
Figure 1. This paper analyses HCI research as problem-solving. Scientific progress in HCI is defined as improvements in our
ability to solve important problems related to human use of computing. Firstly, a subject of enquiry is defined and its improve-
ment potential analysed. Then, a research problem is formulated. The outcome of the research (i.e., the solution) is evaluated
for its contribution to problem-solving capacity defined in terms of five criteria.
clear in almost all definitions of HCI as a field, including that
of the 1992 ACM Curriculum [18]. We therefore propose
adding a constructive problem type. An overview is given in
Figure 2. This typology is orthogonal to the well-known Pas-
teur's Quadrant, which constitutes an attempt to bridge the
gap between applied and basic research by suggesting ‘use-
inspired basic research’ as an acceptable type. In our view,
in HCI, all problems are (somehow) use-inspired and the
quadrant offers little insight.
Empirical Problems
The landscape is replete with empirical problems, across all
HCI venues, from studies of how people use mouseover to
embarrassing experiences with technology and effective
ways of crowdsourcing contributions. Nevertheless, this is
perhaps the most straightforward type to define:
Definition: Empirical research is aimed at creating or elab-
orating descriptions of real-world phenomena related to
human use of computing.
Laudan cites three characteristic subtypes:
1. unknown phenomena
2. unknown factors
3. unknown effects
Qualitative research, ethnography in particular, is an ap-
proach often followed to shed light on novel phenomena. An
example is the 1996 TOCHI article ‘A Field Study of Ex-
ploratory Learning Strategies’ [41], which reported observa-
tions of how users explore software. The constituent factors
of phenomena, however, can be exposed only after the ‘car-
rier’, the phenomenon, has been identified. Consider, for ex-
ample, the paper ‘Distance Matters’ [37]: it catalogues phe-
nomena and factors that affect mediated human-to-human
communication. Finally, after identifying factors, one can
measure and quantify their effects on something of interest.
A common example is evaluative studies wherein statistical
inference is used to quantify the most potent effects. One
could cite fisheye menus here—though there is a great deal
of knowledge about the technique and how to implement it,
a study that evaluated its usability found no benefits of this
technique [20].
Conceptual Problems
Conceptual problems are non-empirical; they involve issues
in theory development in the most general sense. They are
also what Laudan calls second-order problems: their sub-
stance does not pertain to the world directly, unlike empirical
problems. Conceptual problems might involve difficulties in
explaining empirical phenomena, nagging issues in models
of interaction, or seeming conflicts between certain princi-
ples of design. Fitts’ law [45] is perhaps the most well-
known example. It is a statistical model connecting aimed-
movement performance (speed and accuracy) to two proper-
ties of a user interface that designers can affect: distance to
and width of selection areas such as buttons. The research
problem it solves is how performance in aimed movement is
connected to task demands imposed by a UI.
We offer the following, more general definition:
Definition: Work on a conceptual research problem is
aimed at explaining previously unconnected phenomena
occurring in interaction.
Responses to this type of problem include theories, concepts,
methods, principles, and models. Furthermore, Laudan dis-
tinguishes among three characteristic subtypes:
1. implausibility
2. inconsistency
3. incompatibility
We discuss each subtype with well-known examples from
HCI literature. Implausibility means that the phenomenon is
unreasonable, improbable, or lacking an explanation. Con-
sider the 1985 paper in HCI Journal entitled ‘Direct Manip-
ulation Interfaces’ [22], whose authors sought to explain why
GUIs felt more direct and command-language interfaces felt
more indirect. Inconsistency means that a position is incon-
sistent with data, with itself, or with some other position. For
example, empirical research on privacy in HCI led to an ac-
count of privacy as a reciprocal process among two or more
parties to communication [11]. This observation countered
the then-more-common view that privacy is a state or prop-
erty attributable to a technological system. Finally, incompat-
ibility means that two positions have assumptions that cannot
be reconciled. The debate [52] about using throughput (TP)
as a metric for pointing performance falls into this category.
Two scholars proposed two metrics that entailed partially in-
compatible interpretations of the concept and guidance on
how to analyse data.
Constructive Problems
We extend the typology of problems with a third type:
Definition: Constructive research is aimed at producing
understanding about the construction of an interactive arte-
fact for some purpose in human use of computing.
We put emphasis on understanding: the objective is not the
construction itself but the ideas or principles it manifests.
This problem type covers some of the sub-areas of HCI
showing the most vitality at conferences, including interac-
tive systems, interactive applications, interface and sensor
technology, interaction techniques, input devices, user-inter-
Figure 2. The problem-solving view ‘collapses’ research problems
in HCI into three main categories, each with three subtypes.](https://image.slidesharecdn.com/hci-research-as-problem-solving-chi16-160512002517/85/HCI-Research-as-Problem-Solving-CHI-16-presentation-slides-2-320.jpg)
HCI Research as Problem-Solving [CHI'16, presentation slides]
- 1. HCI RESEARCH AS PROBLEM-SOLVING Antti Oulasvirta Aalto University Kasper Hornbæk University of Copenhagen
- 2. A PAPER PRESENTED AT CHI 2016 Paper: http://users.comnet.aalto.fi/oulasvir/pubs/hci-research-as-problem-solving-chi2016.pdf HCI Research as Problem-Solving Antti Oulasvirta Aalto University, Finland Kasper Hornbæk University of Copenhagen, Denmark ABSTRACT This essay contributes a meta-scientific account of human– computer interaction (HCI) research as problem-solving. We build on the philosophy of Larry Laudan, who develops problem and solution as the foundational concepts of sci- ence. We argue that most HCI research is about three main types of problem: empirical, conceptual, and constructive. We elaborate upon Laudan’s concept of problem-solving ca- pacity as a universal criterion for determining the progress of solutions (outcomes): Instead of asking whether research is ‘valid’ or follows the ‘right’ approach, it urges us to ask how its solutions advance our capacity to solve important problems in human use of computers. This offers a rich, generative, and ‘discipline-free’ view of HCI and resolves some existing de- bates about what HCI is or should be. It may also help unify efforts across nominally disparate traditions in empirical re- search, theory, design, and engineering. Author Keywords Human–computer interaction; Problem-solving; Scientific progress; Research problem; Larry Laudan ACM Classification Keywords H.5.m. Information interfaces and presentation (e.g., HCI): Miscellaneous INTRODUCTION The spark for writing this essay comes from feelings of con- fusion, and even embarrassment, arising in describing our field to students and other researchers. What is human–com- puter interaction (HCI) as a field? As numerous ideas and disciplines contribute to HCI, its unique character is elusive. Although HCI is in intellectual debt to many other fields, few would agree that it reduces to them. It has its own subject of enquiry, which is not part of the natural or social sciences. It does not belong to engineering, computer science, or design either. So what is it? The essay has a grand ambition: to develop a conceptually coherent account of the ‘95% of HCI research’. We know of no other paper offering an attempt to address the field as a whole. We are motivated first and foremost by the intellec- tual enigma pertaining to what HCI is: There is no accepted account that would tell how HCI’s numerous approaches contribute to pursuit of shared objectives. In contrast, HCI has been criticised for lack of ‘motor themes, mainstream topics, and schools of thought’ [25] and for being fragmented ‘across topics, theories, methods, and people’ [38]. Conse- quently, some have called for ‘a hard science’ [36], others for ‘strong concepts’ [19] or an ‘inter-discipline’ [3]. These are serious concerns with serious implications for the field. Why bother with a meta-scientific paper at a technical con- ference? Because the stakes are high. Philosophies of science are at worst an impotent topic worthy of hallway conversa- tions. But if the critics are right, our field is seriously crip- pled, from the project level to the larger arenas of research Realpolitik. Lacking a coherent view of what HCI is, and what good research in HCI is, how can we communicate re- sults to others, assess research, co-ordinate efforts, or com- pete? In addition, as we show, philosophical views offer thinking tools that can aid in generating ideas and generally enhance the quality of research. The contribution here lies in describing HCI as prob- lem-solving. An overview is given in Figure 1. The view originates from Larry Laudan’s philosophy of science [28]. Laudan describes scientific progress in terms of two founda- tional concepts: research problem and solution. Laudan's ‘problem’ is not what we mean by the term in ordinary lan- guage. It is defined via inabilities and absences occurring in descriptions; knowledge; or, as often in HCI, constructive so- lutions. For example, a research problem may involve lack of understanding of how colour schemes on a web page af- fect the aesthetic experience of its use. More generally, Lau- dan’s research problem subsumes what we traditionally un- derstand in HCI as a ‘design problem’ but also problems to do with theory and empirical research. Most of our argumentation builds on a concept put forth by Laudan that links problems with solutions: problem-solving capacity. For Laudan, a solution is something special, too. In the above-mentioned case of aesthetic perception of web pages, possible solutions range from descriptions of self-re- ports to models of aesthetic impressions. These solutions change the status of the inabilities and absences but in differ- ent ways. Laudan qualifies this in terms of improvements to problem-solving capacity. This is counter to some traditional notions of progress [28, p. 14]: In appraising the merits of theories, it is more important to ask whether they constitute adequate solutions to signifi- Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for com- ponents of this work owned by others than ACM must be honored. Ab- stracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from [email protected]. CHI'16, May 07-12, 2016, San Jose, CA, USA © 2016 ACM. ISBN 978-1-4503-3362-7/16/05…$15.00 DOI: http://dx.doi.org/10.1145/2858036.2858283 cant problems than it is to ask whether they are ‘true’, ‘cor- roborated’, ‘well-confirmed’ or otherwise justifiable within the framework of contemporary epistemology. With this definition, the benefit of problem-solving is that it allows covering a wider scope of research than previous ac- counts, which have been restricted to certain disciplines, top- ics, or approaches (e.g., research-through-design [53], inter- action criticism [2], usability science [15], or interaction sci- ence [21]). However, because Laudan developed his view with natural and social sciences in mind, he missed design and engineering contributions. Extending Laudan’s typology to propose that research problems in HCI include not only empirical and conceptual but also constructive problems, we present the first typology developed to encompass most rec- ognised research problems in HCI. It is now possible to de- scribe research contributions regardless of the background traditions, paradigms, or methods. The seemingly multi- or, rather, hyper-disciplinary field is—in the end—about solv- ing three types of problem. This reduces the number of di- mensions dramatically when one is talking about HCI. Having built the conceptual foundation, we return to answer four fundamental questions: 1) What is HCI research, 2) what is good HCI research, 3) are we doing a good job as a field, and 4) could we do an even better job? We aim to show through these discussions that Laudan's problem-solving view is not just ‘solutionism’. It offers a useful, timeless, and actionable non-disciplinary stance to HCI. Instead of asking whether research subscribes to the ‘right’ approach, a system is ‘novel’, or a theory is ‘true’, one asks how it advances our ability to solve important problems relevant to human use of computers. Are we addressing the right problems? Are we solving them well? The view helps us contribute to some longstanding debates about HCI. Moreover, we show that the view is generative. We provide ideas on how to apply it as a thinking tool. Problem-solving capacity can be analysed for individual papers or even whole sub-topics and the field at large. It also works as a spring- board for generating ideas to improve research agendas. We conclude on a positive note by arguing that HCI is nei- ther unscientific nor non-scientific (as some have claimed [40]) or in deep crisis [25]. Such views do not recognise the kinds of contributions being made. Instead, on many counts, HCI has improved problem-solving capacity in human use of computing remarkably and continues to do so. However, as we show, these contributions tend to focus on empirical and constructive problem types. In a contrast to calls for HCI to be more scientific [21], interdisciplinary [3], hard [36], soft [9], or rigorous [40], the systematic weakness of HCI is, in fact, our inability to produce conceptual contributions (theo- ries, methods, concepts, and principles) that link empirical and constructive research. THREE TYPES OF RESEARCH PROBLEM IN HCI Our first point is that the key to understanding HCI as prob- lem-solving is the recognition that its research efforts cluster around a few recurring problem types. We effectively ‘col- lapse’ the (apparent) multiplicity of research efforts under a few problem types. This not only simplifies HCI but also transcends some biasing presumptions arising from method- ology, theory, or discipline. One can now see similarities and differences between, say, an observational study of a novel technology and a rigorous laboratory experiment, without being bound by their traditions. In this section, we 1) introduce Laudan’s notion of research problem briefly, 2) extend his typology to cover engineering and design contributions to HCI, and 3) argue that contribu- tions in HCI can be classified via this typology. Laudan originally distinguished only two types of research problem—empirical and conceptual. These are defined in terms of absence or inabilities to understand or achieve some ends. As we argue below, the two types are applicable also to HCI. However, to not let design ‘off the hook’, HCI should cover engineering and design contributions. This aspect is Figure 1. This paper analyses HCI research as problem-solving. Scientific progress in HCI is defined as improvements in our ability to solve important problems related to human use of computing. Firstly, a subject of enquiry is defined and its improve- ment potential analysed. Then, a research problem is formulated. The outcome of the research (i.e., the solution) is evaluated for its contribution to problem-solving capacity defined in terms of five criteria. clear in almost all definitions of HCI as a field, including that of the 1992 ACM Curriculum [18]. We therefore propose adding a constructive problem type. An overview is given in Figure 2. This typology is orthogonal to the well-known Pas- teur's Quadrant, which constitutes an attempt to bridge the gap between applied and basic research by suggesting ‘use- inspired basic research’ as an acceptable type. In our view, in HCI, all problems are (somehow) use-inspired and the quadrant offers little insight. Empirical Problems The landscape is replete with empirical problems, across all HCI venues, from studies of how people use mouseover to embarrassing experiences with technology and effective ways of crowdsourcing contributions. Nevertheless, this is perhaps the most straightforward type to define: Definition: Empirical research is aimed at creating or elab- orating descriptions of real-world phenomena related to human use of computing. Laudan cites three characteristic subtypes: 1. unknown phenomena 2. unknown factors 3. unknown effects Qualitative research, ethnography in particular, is an ap- proach often followed to shed light on novel phenomena. An example is the 1996 TOCHI article ‘A Field Study of Ex- ploratory Learning Strategies’ [41], which reported observa- tions of how users explore software. The constituent factors of phenomena, however, can be exposed only after the ‘car- rier’, the phenomenon, has been identified. Consider, for ex- ample, the paper ‘Distance Matters’ [37]: it catalogues phe- nomena and factors that affect mediated human-to-human communication. Finally, after identifying factors, one can measure and quantify their effects on something of interest. A common example is evaluative studies wherein statistical inference is used to quantify the most potent effects. One could cite fisheye menus here—though there is a great deal of knowledge about the technique and how to implement it, a study that evaluated its usability found no benefits of this technique [20]. Conceptual Problems Conceptual problems are non-empirical; they involve issues in theory development in the most general sense. They are also what Laudan calls second-order problems: their sub- stance does not pertain to the world directly, unlike empirical problems. Conceptual problems might involve difficulties in explaining empirical phenomena, nagging issues in models of interaction, or seeming conflicts between certain princi- ples of design. Fitts’ law [45] is perhaps the most well- known example. It is a statistical model connecting aimed- movement performance (speed and accuracy) to two proper- ties of a user interface that designers can affect: distance to and width of selection areas such as buttons. The research problem it solves is how performance in aimed movement is connected to task demands imposed by a UI. We offer the following, more general definition: Definition: Work on a conceptual research problem is aimed at explaining previously unconnected phenomena occurring in interaction. Responses to this type of problem include theories, concepts, methods, principles, and models. Furthermore, Laudan dis- tinguishes among three characteristic subtypes: 1. implausibility 2. inconsistency 3. incompatibility We discuss each subtype with well-known examples from HCI literature. Implausibility means that the phenomenon is unreasonable, improbable, or lacking an explanation. Con- sider the 1985 paper in HCI Journal entitled ‘Direct Manip- ulation Interfaces’ [22], whose authors sought to explain why GUIs felt more direct and command-language interfaces felt more indirect. Inconsistency means that a position is incon- sistent with data, with itself, or with some other position. For example, empirical research on privacy in HCI led to an ac- count of privacy as a reciprocal process among two or more parties to communication [11]. This observation countered the then-more-common view that privacy is a state or prop- erty attributable to a technological system. Finally, incompat- ibility means that two positions have assumptions that cannot be reconciled. The debate [52] about using throughput (TP) as a metric for pointing performance falls into this category. Two scholars proposed two metrics that entailed partially in- compatible interpretations of the concept and guidance on how to analyse data. Constructive Problems We extend the typology of problems with a third type: Definition: Constructive research is aimed at producing understanding about the construction of an interactive arte- fact for some purpose in human use of computing. We put emphasis on understanding: the objective is not the construction itself but the ideas or principles it manifests. This problem type covers some of the sub-areas of HCI showing the most vitality at conferences, including interac- tive systems, interactive applications, interface and sensor technology, interaction techniques, input devices, user-inter- Figure 2. The problem-solving view ‘collapses’ research problems in HCI into three main categories, each with three subtypes.
- 3. SHAPE OUR CAREERS AND OUR FIELD BELIEF SYSTEMS Not well- articulated Not internally consistent Not value-free
- 4. “Novelty is important in HCI research” “A good paper reports implications to practitioners” “HCI should be more scientific” “Cognitive modeling is passe”
- 5. “What should I research?” “What is good HCI?” “Is this paper good?” “Is HCI progressing?” mportant in HCI research” orts titioners” re scientific” modeling is passe” Thinking
- 6. Assessment criteria Goal-setting Field at should I research?” at is good HCI?” his paper good?” CI progressing?” Practices inking
- 7. from Blackwell, 2015 ‘Hyper-disciplinarity’ SYSTEMIC TROUBLE IN HCI? HCI “fragmented across topics, theories, methods, and people“ Olson & Olson, 2000 Liu et al., 2014 The ‘big hole in HCI’
- 8. A UNIFYING CONCEPT PROBLEM-SOLVING CAPACITY “Instead of asking whether research is ‘valid’ or follows the ‘right’ approach, it urges us to ask how its solutions advance our capacity to solve important problems in human use of computers.“
- 9. MORE THAN JUST A LENS ’95% OF HCI RESEARCH’ Our identity as a field Inclusive without being naive Generate new ideas, steer thinking
- 10. EXISTING ACCOUNTS CONSTRAIN RESEARCH ‘Hard science’ ‘In the wild’ ‘Inter-discipline’ ‘Waves’ ‘Epochs’ ‘Novelty’ ‘Benefit’ ‘Design implications’ Problem OutcomeApproach
- 11. A NEW UNIT OF ANALYSIS Problem OutcomeSolution
- 12. LARRY LAUDAN In appraising the merits of theories, it is more important to ask whether they constitute adequate solutions to significant problems than it is to ask whether they are ‘true’, ‘corroborated’, ‘well-confirmed’ or otherwise justifiable within the framework of contemporary epistemology. ‘Progress and its Problems’ 1978
- 13. ‘RESEARCH PROBLEM’ “…not what we mean by the term in ordinary language. It is defined via inabilities and absences occurring in descriptions; knowledge; or, as often in HCI, constructive solutions.” design problem
- 14. JUST THREE PROBLEM TYPES “One can now see similarities and differences between, say, an observational study of a novel technology and a rigorous laboratory experiment, without being bound by their traditions.” COLLAPSING MULTI-DISCIPLINARITY
- 15. THREE PROBLEM TYPES IN HCI Research problems in HCI Empirical Unknown phenomena Unknown factors Unknown effects Conceptual Implausibility Inconsistency Incompatibility Constructive No known solution Partial solution Inability to deploy/implement OVERVIEW
- 16. PROBLEM-SOLVING CAPACITY Significance Addresses a problem important to our stakeholders Effectiveness Solves essential aspects of it Transfer and transfers to other problem- instances. Efficiency …and with little use of resources … with high reliability & validity Confidence DEFINITION
- 17. Tangible Bits Ishii & Ulmer, 1997 Constructive Conceptual TWO CLASSIC HCI CONTRIBUTIONS Fitts’ Law ILLUSTRATIVE EXAMPLE
- 18. EXAMPLES OF CONSTRUCTIVE CONTRIBUTIONS Problem: Significance Solution: Capacity Tangible Bits All input Pointing tasks Vision, demonstrators Model Fitts’ Law
- 19. HCI RESEARCH AS PROBLEM- SOLVING ZOOMING OUT Achieved progress Research problems Desired problem-solving capacities Description of a field
- 20. In a paper, often more than one type present All problem types and capacities present High tolerance for risk ‘Significance’ often shaped by society/industry Achieved progress Research problems Desired problem-solving capacities SOLVING HCI in general
- 21. In a paper, often more than one type present All problem types and capacities present High tolerance for risk ‘Significance’ often shaped by society/industry Capacities often ambiguous CHI’15 Best PapersMostly empirical and constructive Achieved progress capacities Problems well- described We miss conceptual work relating empirical and constructive capacities
- 22. EXPLAINING ‘THE BIG HOLE IN HCI’ Problem-solving capacity Time Paradigm-shifting ‘Leapfrogging’
- 23. IN OUR BELIEF SYSTEMS “Novelty is important” “HCI is about design” “HCI should be scientific” EXPOSING FLAWS Ignores the types of problems and capacities we need in HCI. Ignores either solution or problem.
- 24. A GENERATIVE CONCEPT PROBLEM-SOLVING CAPACITY Example in paper Fitts’ law
- 25. A DISCIPLINE-FREE VIEW OF HCI RESEARCH SUMMARY Embraces variety without naivety Shows that we need work on conceptual problems Shows that we do have significant progress You can use this to generate and refine research ideas A single concept; much richer than “solutionism”
- 26. HCI Research as Problem-Solving / CHI’16 COMMON OBJECTIONS 1. It’s solutionism 2. It does not establish the boundaries of HCI with other disciplines 3. It ignores the role of art in HCI 4. It ignores the role of curiosity 5. Some topics are too subjective to be problem-solving 6. HCI, especially in design, is inherently messy 7. The view is iffy and leads to lot of handwaving 8. It ignores HCI’s impact on society SCOPE AND LIMITATIONS
- 27. We thank Susanne Bødker, Stuart Reeves, Barry Brown, Antti Salovaara, Giulio Jacucci, Vassilis Kostakos, Pierre Dragicevic, Andrew Howes, and Pertti Saariluoma. Instead of asking whether research is ‘valid’ or follows the ‘right’ approach, it urges us to ask how its solutions advance our capacity to solve important problems in human use of computers. HCI RESEARCH AS PROBLEM-SOLVING