Situated action and ethnomethodology: the plan is an account, not the mechanism

The situated-action critique, the subject of this fifth of ten units, is usually heard as the claim that people do not plan. Its real claim is narrower. A plan is something a person produces from within a course of action, and uses as one resource among many, not the machinery that generates the action (Suchman, 1987). If that is right, a system fails on exactly the non-routine occasions that matter. It requires its user to state a plan before it will help, and reads any departure from it as error. Those are the occasions where plan and situation come apart. The strongest reply, that procedure models remain valid for routine work, is real but narrower than the critique, so the dispute is unsettled.

Plans are a resource for action, not its mechanism

The critique begins in ethnomethodology. Garfinkel (1967) studied members' own methods for producing and recognising orderly action. He drew the consequence that social order is accomplished in the doing rather than given in advance. A description of order cannot then be read as a specification of the machinery that produced it. Suchman (1987) carried the argument into interaction design with a contrast between two ways of navigating. A European navigator plots a course before departure and relates every move to it. A Trukese navigator begins with an objective, steers by wind, waves, tide and stars as they arise, and can point to the objective but not describe a course. The claim is not about two cultures but about all activity. Even the most analytic action is concrete and embodied, so purposeful action is inevitably situated action. Plans are then a weak resource for what is primarily ad hoc activity. They are vague in advance, and in retrospect they filter out the particularity of what was done (Suchman, 1987).

Competence moved with the setting, not the person

The critique would be a preference without evidence, and Lave (1988) supplied it. The Adult Math Project followed the same people through arithmetic in everyday life and on a test. It found a discontinuity within the same heads. On best-buy problems the shoppers were correct 98 per cent of the time in the supermarket, using the prices and packages in front of them. They averaged 59 per cent on school-style paper-and-pencil tests, and a best-buy simulation scored 93 per cent (Lave, 1988, p. 56). Their procedures differed too. In the aisle they decomposed numbers from the largest part down, rounded and used the environment as a calculating device. The tests rewarded carrying and borrowing. Lave reports the same pattern from Carraher, Carraher and Schliemann (1985). Market sellers in Recife were correct 99 per cent of the time in the market, against 74 per cent on a paper test with identical problems. What arithmetic people did, and how well they did it, moved with the setting. Brown, Collins and Duguid (1989) carried the claim into learning. They argued that knowledge is situated, a product in part of the activity, context and culture in which it is used.

The help system inferred a plan the user had not declared

Suchman's (1987) own evidence is an analysis, recorded on video, of an intelligent help system on a photocopier. The design assumed that a user's purpose would suffice as context. The machine inferred the plan behind whatever action it could detect, and used that plan to read the next action. The recordings showed the relation between intention and action to be too weak for the inference. When users meant "what happened?" or "how do I do this?", the system could answer "describe the options of this display". Users then misread the machine by imposing conventions from human conversation. A screen reader widens that inference gap. Suchman separated the trouble into a false alarm, where a user believes an error occurred when none did, and a garden path, where an error goes unnoticed. She noted the system had no access to the user's situation to detect either. A system that requires people to declare their plan before it will help, and treats deviation as error, fails exactly on the occasions that matter.

Ethnomethodology is a study policy, not a specification

If the critique is right, can it be turned into a method for building systems? The literature answers with difficulty rather than a recipe. Button (1990) argued against conflating the analysis of conversation with computational modelling. He distinguished rules people can be shown to orient to, embedded within action, from rules held to be interior mental machinery that stands behind it. A model built on the second reading misconstrues what the first describes. Dourish and Button (1998) made the most ambitious attempt to cross the seam. They argued that ethnomethodology could take a foundational place in system design, rather than serving only as a resource for requirements elicitation. They were candid that a settled way of doing so was lacking. The programme also produced demonstrations of an accountable system, made with Dave Curbow. A folder being copied to a file server is shown to the user as the copy proceeds. When a user asks for more detail, the system records the preference and supplies it the next time that situation arises. The system's own working becomes visible rather than hidden. Anderson (1994) frames the same tension from the design side. Designers turned to ethnography chiefly for the specification of end-user requirements. Its value must be weighed against the objection that it returns descriptions rather than requirements a specification can use.

Contested question: is the critique decisive, and about what?

The literature disagrees in two directions at once. The situated-action position holds that the critique is decisive for design. Its evidence is Lave's (1988) discontinuity, Suchman's (1987) copier breakdowns and the generalisation to learning by Brown, Collins and Duguid (1989). Greeno (1998) states the perspective for a psychology audience, treating an intact activity system, individual agents included, as the unit of analysis. What would settle it is a demonstration that a system requiring a stated plan degrades performance on non-routine tasks while leaving routine performance intact. The degradation must follow from the plan requirement, not from novelty.

The symbolic reply is the strongest case against the thesis. Vera and Simon (1993) argue that symbolic systems already perform well in temporally demanding tasks. The systems usually cited as exemplifying situated action are themselves representational, they add, and its goals can be met only within a symbolic framework. Their own illustration of a plan in use is a canoeist shooting rapids. The plan is an approximate path through the water, as broad as the river allows. It takes the main currents and avoids the obvious obstacles, and execution adjusts it continuously to what the current offers. A plan, on that reading, is what reduces the number of occasions on which an emergency demands situated action. Suchman's (1993) reply is that this reinstates the plan-based assumptions under dispute. She treats plans as conceptual and rhetorical devices for making sense of action, not as its generative mechanism. The exchange turns on what is being claimed, a claim about mechanism or a claim about the sufficiency of plans.

The third position moves the dispute to education and reframes it as one about form. Anderson, Reder and Simon (1996) review four claims of situated learning: that action is grounded in its concrete situation, and that knowledge does not transfer. The remaining two are that abstraction is of little use, and that instruction must be in complex social settings. They cite the literature to argue that each is overstated. They then hold that the situative and cognitive perspectives differ in form rather than substance and can be reconciled (Anderson, Reder and Simon, 1997). Greeno (1998) answers that a framework can be broader without being merely stylistic, which leaves the contest unresolved. None of the three is decidable from the present record, because the critiques and replies are not always about the same proposition. A comparison that fixed the outcome and reported whether any breakdown followed from the plan requirement itself would separate them.

The two sides do not carry the same kind of evidence. Lave's (1988) within-person comparison and Suchman's (1987) video analysis are primary data. Vera and Simon (1993) restate a position, and Anderson, Reder and Simon (1996) review other people's claims. The comparison is confounded in one respect. The shoppers' scores track the form in which the problem was posed as much as the setting it was posed in. While the dispute is open, a designer should build the plan-recording variant beside the plan-demanding one. A breakdown counts as evidence only if the requirement produced it. The strongest objection is that no study here compares a plan-requiring system with an otherwise identical one, so the copier breakdowns cannot be attributed to the plan requirement.

Where the argument is heading

Suchman (2007) restates the situated-action argument. She adds an account of how persons and machines are enacted as similar or different in practice. That moves the question from the status of human-like machines to how the line between them is drawn. Heft (2024) extends the lineage differently, reading behaviour settings into situated action. Barker and Wright (1955) tracked the daily behaviour of some twenty children in a small midwestern town of roughly 700 people. They found that actions which looked unpredictable at the level of the individual were reliably context-dependent once their place was taken into account. Heft argues that settings and complexity theory give the level at which situated action should be analysed. He asks for more attention to the individual's improvisational action. The unit's thesis stands: the plan remains an account produced within action, the routine-task reply answers a narrower question, and the dispute stays open.

The sources this unit rests on

• Anderson, J. R., Reder, L. M. and Simon, H. A. (1996). Situated Learning and Education. Educational Researcher 25(4), 5-11. doi:10.3102/0013189x025004005 Reviews the four central claims of situated learning and cites empirical work to argue each is overstated. The cognitive-science counter-attack in the terms of education.
• Anderson, J. R., Reder, L. M. and Simon, H. A. (1997). Situative Versus Cognitive Perspectives: Form Versus Substance. Educational Researcher 26(1), 18-21. doi:10.3102/0013189x026001018 Replies to Greeno, holding that the situative and cognitive perspectives differ in form rather than substance and can be reconciled. It supplies the terms of the exchange.
• Anderson, R. J. (1994). Representations and Requirements: The Value of Ethnography in System Design. Human-Computer Interaction 9(2), 151-182. doi:10.1207/s15327051hci0902_1 Examines why designers turned to ethnography, chiefly for specifying end-user requirements, and weighs its value against the objection that it returns descriptions rather than requirements.
• Barker, R. G. and Wright, H. F. (1955). Midwest and Its Children: The Psychological Ecology of an American Town. Harper and Row. doi:10.1037/10027-000 The field study Heft (2024) reports: a small midwestern town of about 700 people, where the daily behaviour of some twenty children was tracked in its behaviour settings.
• Brown, J. S., Collins, A. and Duguid, P. (1989). Situated Cognition and the Culture of Learning. Educational Researcher 18(1), 32-42. doi:10.3102/0013189x018001032 Argues that knowledge is situated, a product of the activity and culture in which it is used, so teaching that abstracts knowledge from use limits itself. It carries the critique into learning.
• Button, G. (1990). Going Up a Blind Alley: Conflating Conversation Analysis and Computational Modelling. In Luff, P., Gilbert, N. and Frohlich, D. (eds), Computers and Conversation. Academic Press, 67-90. doi:10.1016/B978-0-08-050264-9.50009-9 Distinguishes rules people orient to, embedded within action, from rules taken as interior mental machinery behind it, and argues that modelling conversation as their execution misconstrues the analysis.
• Carraher, T. N., Carraher, D. W. and Schliemann, A. D. (1985). Mathematics in the streets and in schools. British Journal of Developmental Psychology 3(1). doi:10.1111/j.2044-835x.1985.tb00951.x Reports the study Lave takes second-hand: market sellers in Recife solved the same problems 99 per cent correctly in the market and 74 per cent on a paper test.
• Dourish, P. and Button, G. (1998). On "Technomethodology": Foundational Relationships Between Ethnomethodology and System Design. Human-Computer Interaction 13(4), 395-432. doi:10.1207/s15327051hci1304_2 Argues that ethnomethodology could take a foundational place in system design rather than serving only requirements elicitation. The attempt, and its acknowledged difficulty, tests whether the critique can become a method.
• Garfinkel, H. (1967). Studies in Ethnomethodology. Prentice-Hall. (Chapter 'The Program of Ethnomethodology', reprinted 2019, Harold Garfinkel: Parsons' Primer, 327-338; doi:10.1007/978-3-476-04815-8_3) Sets out the study of members' own methods for producing and recognising orderly action, making social order an accomplishment rather than a given.
• Greeno, J. G. (1998). The situativity of knowing, learning, and research. American Psychologist 53(1), 5-26. doi:10.1037/0003-066x.53.1.5 States the situative perspective for a psychology audience, treating an intact activity system, individual agents included, as the unit of analysis.
• Heft, H. (2024). Behaviour settings, situated action and complexity theory. Philosophical Transactions of the Royal Society B: Biological Sciences 379(1910), 20230284. doi:10.1098/rstb.2023.0284 Reads Barker's behaviour settings back into situated action, reporting that children's actions which looked unpredictable individually became reliably context-dependent once their place was taken into account.
• Lave, J. (1988). Cognition in Practice: Mind, Mathematics and Culture in Everyday Life. Cambridge University Press. doi:10.1017/cbo9780511609268 The Adult Math Project followed the same people through everyday and test-like arithmetic; on best-buy problems shoppers scored 98 per cent correct in the supermarket against 59 per cent on school-style tests.
• Suchman, L. (1993). Response to Vera and Simon's Situated Action: A Symbolic Interpretation. Cognitive Science 17(1), 71-75. doi:10.1207/s15516709cog1701_5 Argues that Vera and Simon's reinterpretation reinstates the plan-based assumptions under dispute, treating plans as conceptual and rhetorical devices for making sense of action.
• Suchman, L. A. (1987). Plans and Situated Actions: The Problem of Human-Machine Communication. Cambridge University Press. https://archive.org/details/planssituatedact0000such Contrasts a European navigator who plots a course with a Trukese navigator who steers by wind, waves and stars, and analyses video of first-time users of a photocopier help system whose plan-inference broke down repeatedly.
• Suchman, L. A. (2007). Human-Machine Reconfigurations: Plans and Situated Actions, 2nd edition. Cambridge University Press. doi:10.1017/cbo9780511808418 Restates the situated-action argument and adds an account of how persons and machines are enacted as similar or different in practice, the author's later revision.
• Vera, A. H. and Simon, H. A. (1993). Situated Action: A Symbolic Interpretation. Cognitive Science 17(1), 7-48. doi:10.1207/s15516709cog1701_2 Contests the situated-action claims, arguing that symbolic systems already perform well in temporally demanding tasks, that systems cited as situated are themselves representational, and that its goals can be met symbolically.