Paradigms and waves: what the field thinks it is

A field that cannot agree on what it is cannot agree on what counts as good work in it. Human-computer interaction is described as a branch of human factors, as an applied cognitive science, as the study of situated practice, and now as the study of people and machines entangled beyond separation. Each description carries a different answer to a question a designer needs. What would show this claim to be wrong?

This unit argues that the incompatible descriptions are not a failure of maturity that more evidence will repair. They are the field's structure. A piece of work can be placed only by asking which description's test of truth it was written to pass. Placement comes before judgement, because the standards differ before the results do. The next unit takes one style the field treated as settled and asks which test it was held to.

A paradigm is a claim about what counts as knowledge

Harrison, Tatar and Sengers (2007) named three paradigms: human factors, classical cognitivism, and a third they call phenomenologically situated. Each is a world view with its own metaphor of interaction. Each brings certain phenomena to the centre and marks others marginal. The three, they write, have "their own forms of knowledge creation and criteria for what constitutes knowledge". A paradigm is not a topic area. It is a set of commitments about what a good question is, which methods answer it, and what validation a claim requires.

Following Agre's account of generative metaphors rather than Kuhn's absolutism, they argue that paradigms can exist side by side without one being shown wrong.

Three faces grew apart in different communities

Grudin (2005) supplies the mechanism behind the pluralism. He examines three research foci that all sit under the field's name: computer operation, information systems management, and discretionary use. The field has not coalesced. Many of its researchers reside in other fields, and forces have kept the three foci apart.

The three are not categories only. The first extended human factors to computing, where use was not a choice and the goals were shorter training and fewer errors. The second grew from business computing on mainframes in the 1960s, with a hands-off user. The third arose with minicomputers and home computers in the 1980s. One name, three communities, three kinds of user.

If that history holds, the descriptions of the field differ because their communities differ. A paradigm label then predicts where a paper was published at least as well as it predicts what the paper measures. That last step is an inference rather than Grudin's claim.

The first move was argued, not measured

The clearest early case of a paradigm being argued into place is Bannon's (1991) chapter. He argues that psychology, in its human factors form, had neglected the values of the people involved and their motivation in the work setting. The two terms in his title name two perceptions of a person. Human factors connotes a passive, fragmented, depersonalised and unmotivated individual. Human actors connotes an active and controlling one. He adds that the field's focus is biased towards technology, so people appear as users, and as naive users.

What the chapter does not report is a comparison of the framings. It re-describes the person in the design process and argues that the description changes what gets designed. Activity theory arrived in design around the same years from a European tradition, recorded by Kuutti (1996). Harrison, Tatar and Sengers (2007) record its cost: activity theory is often incorporated only to create accounts of an existing situation, so its relevance to design and evaluation is treated as limited.

Activity theory is a third position, not a footnote to the other two

The taxonomy becomes harder to use once a framework arrives that is neither cognitivist nor the ethnography of a setting. Kuutti (1996) records the debate that opened the space. Information processing cognitive psychology, the mainstream framework for interaction research, had come under increasing criticism for problems in applying it. Kuutti presents activity theory as an alternative for research and design.

The framework's founding illustration makes the unit of analysis concrete. Leontiev's example, described by Kaptelinin (2005), is a beater in a primeval collective hunt, employed to frighten animals towards hunters waiting in ambush. His own goal is the opposite of his motive, which is food, and the act makes sense only through his relation to the others in the hunt. Leontiev uses this to separate an action, which answers to a goal, from an activity, which answers to a motive. A design that models only the steps of a task has modelled the actions and missed the activity.

Kaptelinin and Nardi (2017) restate that framework two decades later. They begin from the claim that the relationship between people and technology has always interested activity theory, which follows from its focus on mediation and tools. On their account a tool stands between a person and the object of the work rather than being a surface with properties to be optimised. On that reading, offered here as interpretation rather than quotation, activity theory is a third position rather than a variant of either paradigm above.

The waves are a different periodisation, and they cut across the paradigms

Bodker's (2006) waves are a second way of describing the field, and they are not the same cut as the paradigms. Her third wave broadens and intermixes use contexts.

Bodker (2015) sets out the waves ten years on. The first wave was model-driven, drawn from cognitive science and human factors, and studied through guidelines and systematic testing. The second wave addressed groups, work settings and communities of practice, drawing on situated action and activity theory, with participatory design workshops and contextual inquiry in the toolbox. The third wave saw technology spread into homes and everyday life, and it challenged the second wave's values. Homes and everyday life are where users who are older, disabled or new to the interface sit, and the first wave's measures were built for an operator at work.

Two things bear on this unit's thesis. Bodker reports that Harrison, Tatar and Sengers reached an almost parallel analysis to her 2006 keynote, so two periodisations of one field appeared independently and within a year of each other. She also asked whether the field had become better at studying experience-based use, and she treated the answer as open.

Epistemology, not method, is what the dispute is about

Harrison, Sengers and Tatar (2011) sharpen the disagreement by moving it from methods to knowledge. They identify a cluster of research cutting across the usual topic areas that shares an epistemological orientation aligned with standpoint epistemology. It acknowledges the situatedness of both the users studied and the analysts studying them. They argue that this successor science already exists within the field, though its practitioners may not recognise it as such.

Rogers (2004) gives the pluralism its bill. The theories imported into interaction research broadened the field's discourse while having limited impact on the practice of interaction design.

Contested question: do the descriptions coexist or compete?

Three positions are on the table. Harrison, Tatar and Sengers (2007) hold the coexistence position deliberately. Following a metaphor account rather than an absolutist one, they can say that the paradigms need not be reconciled. No ranking is intended, and components of work take their meaning from the paradigm they sit in.

Harrison, Sengers and Tatar (2011) hold the competing-standards position in its strongest form. Their third paradigm is a successor science with its own legitimation, and the practitioner must declare the standpoint from which a claim is made. Grudin (2005) undercuts both. The evidence behind the three positions is unequal. Harrison, Tatar and Sengers (2007) rest their third paradigm on a count of 30 papers out of 151 and on their own illustrative cartoon. Grudin (2005) is archival history. Harrison, Sengers and Tatar (2011) propose a standard rather than demonstrate one.

The disagreement turns on what would count as evidence, and none of these sources stages the test. If coexistence is right, a paradigm declared before a study is run should predict the method chosen and the criteria by which the result is judged. Those criteria should hold across settings. If the competing-standards reading is right, a study judged by two paradigms should fail one of them for reasons the other accepts, and a reviewer would record that failure. Bodker's (2015) question about methodical progress is the nearest available proxy. The strongest counter to this unit's thesis is that its central move is always available: any paper can be assigned a paradigm after the fact, so the claim that the field's structure is plural resists the test it sets for others.

A paradigm decides what its authors fail to observe

The one case here where an assumption was caught out by observation is the paperless office. Sellen and Harper (2002) studied paper use in offices and found that paper persisted because it served the tasks at hand. Those tasks were flexible navigation through a document, cross-referencing of several documents at once, annotation, and the interweaving of reading and writing.

Their reported conclusion is that an attempt to remove paper failed because the ways people worked did not change. The change had to be made in the work processes, with the reduction of paper a consequence rather than a target. The failure is a paradigm failure rather than a technical one.

Bannon (2011) gives that movement a name. He describes the field's trajectory as a sequence of requests: evaluate this interface, help us design it, find out what users need, look at an area of life and find something interesting in it. He concludes that the subject is now human activities mediated by computing.

While it is open, the designer's version of the placement is to name the account of interaction an interface assumes and to say what would show that account wrong in use. That is a claim a design can be held to, and it does not require the field to settle which paradigm is right.

Where the argument is heading

The next move in the literature is a further paradigm rather than a further wave. Frauenberger (2019) argues that intimacy with digital technologies generates ontological uncertainty, epistemological diffusion and ethical difficulty. He holds that the response is to evolve the research paradigm from post-humanism and relational ontologies such as actor-network theory and agential realism.

Bodker (2015) is the sceptic in the room. Asked about a fourth wave, she answered that the field is "in the middle of a chaos of multiplicity in terms of technologies, use situations, methods, and concepts", leaving what lies beyond to others. What the literature points to is a placement discipline: saying which paradigm a piece of work is in, and which test it is written to pass, at the point of publication.

The sources this unit rests on

• Harrison, S., Tatar, D. and Sengers, P. (2007). The three paradigms of HCI. alt.chi session at CHI 2007, San Jose. https://people.cs.vt.edu/srh/Downloads/HCIJournalTheThreeParadigmsofHCI.pdf Named three paradigms, each with its own metaphor, questions, methods and criteria of knowledge, and counted 30 of 151 CHI 2006 papers as situated work. Ranked none.
• Grudin, J. (2005). Three faces of human-computer interaction. IEEE Annals of the History of Computing 27(4), 46-62. doi:10.1109/MAHC.2005.67 Traced three research foci, computer operation, information systems management and discretionary use, and reported that the field has not coalesced.
• Bannon, L. J. (1991). From human factors to human actors: the role of psychology and human-computer interaction studies in system design. Design at Work (Greenbaum and Kyng, eds.), Lawrence Erlbaum, 25-44. https://www.taylorfrancis.com/chapters/edit/10.1201/9781003063988-3/human-factors-human-actors-role-psychology-human-computer-interaction-studies-system-design-liam-bannon Argued that human factors psychology neglected the values and motivation of the people involved, contrasting a passive, depersonalised person with an active one.
• Kuutti, K. (1996). Activity theory as a potential framework for human-computer interaction research. Context and Consciousness (Nardi, ed.), MIT Press, 17-44. doi:10.7551/mitpress/2137.003.0006 Records that information processing cognitive psychology had been criticised for serious problems in applying it in research and design.
• Kaptelinin, V. and Nardi, B. (2017). Activity theory as a framework for human-technology interaction research. Mind, Culture, and Activity 25(1), 3-5. doi:10.1080/10749039.2017.1393089 Restates the framework, whose interest in technology follows from its focus on mediation and tools. Programmatic, not empirical.
• Rogers, Y. (2004). New theoretical approaches for human-computer interaction. Annual Review of Information Science and Technology 38(1), 87-143. doi:10.1002/aris.1440380103 Surveyed the theories imported into the field: they broadened its discourse yet barely touched design practice.
• Harrison, S., Sengers, P. and Tatar, D. (2011). Making epistemological trouble: third-paradigm HCI as successor science. Interacting with Computers 23(5), 385-392. doi:10.1016/j.intcom.2011.03.005 Argue that a research cluster shares a standpoint epistemology acknowledging the situatedness of users and analysts, and that this successor science exists unrecognised.
• Bodker, S. (2006). When second wave HCI meets third wave challenges. Proceedings of the 4th Nordic Conference on Human-Computer Interaction (NordiCHI '06), 1-8. doi:10.1145/1182475.1182476 Described the second wave as work-oriented and the third as broadening and intermixing use contexts.
• Bodker, S. (2015). Third-wave HCI, 10 years later: participation and sharing. Interactions 22(5), 24-31. doi:10.1145/2804405 Revisits the 2006 keynote and restates the three waves. Reports that Harrison, Tatar and Sengers reached an almost parallel analysis, and leaves open whether the field became better at studying experience-based use.
• Bannon, L. (2011). Reimagining HCI: toward a more human-centred perspective. Interactions 18(4), 50-57. doi:10.1145/1978822.1978833 Argues that the field's subject has become human activities mediated by computing, and calls for a human-centred reformulation.
• Frauenberger, C. (2019). Entanglement HCI: the next wave? ACM Transactions on Computer-Human Interaction 27(1), 1-27. doi:10.1145/3364998 Argues that entanglement with digital technologies challenges the field's foundations, and that the paradigm should be evolved from relational ontologies.
• Sellen, A. J. and Harper, R. H. R. (2002). The Myth of the Paperless Office. MIT Press. https://www.microsoft.com/en-us/research/publication/myth-paperless-office/ Studied paper use in offices and reported that paper served the tasks at hand. Removing paper failed because the ways people worked did not change; the note rests on a published description and review, not the text.
• Kaptelinin, V. (2005). The object of activity: making sense of the sense-maker. Mind, Culture, and Activity 12(1), 4-18. doi:10.1207/s15327884mca1201_2 Describes and re-reads Leontiev's example of the beater in a primeval collective hunt, separating an action, which answers to a goal, from an activity, which answers to a motive. This unit cites it for the illustration, not for a study of computer use.