November 7, 2003                               Situated Action                         Michelle Cooper

http://ella.slis.indiana.edu/~micooper/L642/SituatedAction.htm

 

Summary

Situated action was first introduced in 1987 in Lucy Suchman's book, Plans and Situated Actions: the problem of human-machine communication.  Since then there have been many papers written about situated action and its role in HCI.  Listed below are just a few of the different views of situated action.

 

Situated action "stresses the knowledgeability of actors and how they use common-sense practices/procedures to produce, analyze and make sense of one another's actions and their local or situated circumstances" (Doerry, 1995).

 

"The term situated actions emphasizes the interrelationship between an action and its context of performance." (Chen & Rada)

 

"Usability First" gives the definition of situated action as "the notion that people's behavior is contextualized, i.e. the situation is a very important factor in determining what people will do. In the extreme view, this is the idea that you can't generalize and predict people's behavior from one situation to the next. Thus, this suggests an approach to usability which says to understand each user's or, more commonly, each organization's specific and detailed needs in designing software for them by carefully examining how they work and how situational and organizational factors fit into that process."

 

"Situated activity is not a kind of action, but the nature of animal interaction at all times, in contrast with most machines we know. This is not merely a claim that context is important, but what constitutes the context, how you categorize the world, arises together with processes that are coordinating physical activity. To be perceiving the world is to be acting in it--not in a linear input-output relation (actñobserveñchange)--but dialectically, so that what I am perceiving and how I am moving co-determine each other" (Clancey, 1993).

 

The four views of situated action above differ slightly in their explanations, but all drive the point that people's actions are influenced by the context of their specific situation.  The last view goes a bit further by drawing on how a users perceptions of the situation and specific actions are continually working together to determine the next step.  Of course, Lucy Suchman described it best herself by stating that every course of action is highly dependent upon its material and social circumstances focusing on moment-by-moment interactions between actors, and between actors and the environments of their action (Suchman, 1987).

 

Motivation

Lucy Suchman introduced the term situated action in her book, Plans and Situated Actions: the problem of human-machine communication, as a way to reformulate the problem of purposeful action. "To designate the alternative that ethnomethodology suggests – more a reformulation of the problem of purposeful action, and a research programme, than an accomplished theory – I have introduced the term situated action" (Suchman, 1987).  She also clarifies that her intentions are not to produce more formal models, "I have proposed an alternative approach drawn from recent developments in the social sciences, principally anthropology and sociology.  The aim of research, according to this approach, is not to produce formal models of knowledge and action, but to explore the relation of knowledge and action to the particular circumstances in which knowing and acting invariably occur" (Suchman, 1987).

 

History and Background

In a 1999 interview with Dialog on Leadership, Lucy Suchman explains that around the time she published her book she was becoming involved with the AI community and with the idea of intelligent interaction interfaces.  She got involved in a project at Xerox PARC "that led to the study that became my dissertation and then was published as Plans and Situated Actions" (Scharmer, 1999).  That project was about a photocopier that was having trouble in the marketplace because it was too difficult to use.  Her colleagues at Xerox PARC built "an expert system that would replace the various instructions that went with the machine" (Scharmer, 1999). Suchman's study ended up being about the different kinds of assumptions the AI community was developing about actions and communication which were being used to develop these expert systems and intelligent interfaces.  The goal of the study was to question these assumptions and create a different approach.  It was also "building on a similar kind of argument about the relationship between instructions and the actual work of operating a machine" (Scharmer, 1999). 

 

Explanation

Simply, situated action is how people act in a situation. 

 

Suchman (1987) explains it in more detail by signifying what it is and what it is not,

Rather than attempting to abstract action away from its circumstances and represent it as a rational plan, the approach is to study how people use their circumstances to achieve intelligent action.  Rather than build a theory of action out of a theory of plans, the aim is to investigate how people produce and find evidence for plans in the course of situated action. More generally, rather than subsume the details of action under the study of plans, plans are subsumed by the larger problem of situated action.

People often have plans of action mapped out in their heads, but may need to change that plan depending on what is actually happening in a specific situation.  They use their embodied skills or past experiences to get them through the situation. Suchman (1987) gives an example of going over the falls in a canoe:

in planning to run a series of rapids in a canoe, one is very likely to sit for a while above the falls and plan one's decent.  The plan might go something like "I'll get as far over to the left as possible, try to make it between those two large rocks, then backferry hard to the right to make it around that next bunch." A great deal of deliberation, discussion, simulation, and reconstruction may go into such a plan.  But, however detailed, the plan stops short of the actual business of getting your canoe through the falls.  When it really comes down to the details of responding to currents and handling a canoe, you effectively abandon the plan and fall back on whatever embodied skills are available to you.

This situation is significant to situated action because it shows how people build detailed plans of action before performing a task.  This also shows how the plan ends up changing once the person is actually performing the task.  Suchman believes that people construct their plan as they go along in the situation, creating and altering their next move based on what has just happened.  People can attempt to make a plan, but their situation will ultimately determine what actual plan of action they make. 

 

In other instances, people may act according to the situation ad hoc.

A participant in the Weight Watchers program had the task of fixing a serving of cottage cheese that was to be three-quarters of the two-thirds cup of cottage cheese the program normally allotted. To find the correct amount of cottage cheese, the dieter, after puzzling over the problem a bit, filled a measuring cup two-thirds full of cheese, dumped it out on a cutting board, patted it into a circle, marked a cross on it, scooped away one quadrant, and served the rest (Nardi, 1995).

Suchman (1987) states that "we generally do not anticipate alternative courses of action, or their consequences, until some course of action is already under way."

 

These ideas of acting according to what is happening are the embodiment of situated action.  The idea is to look at how users interact with machines according to what is happening.  This can help designers plan for specific activities when users are in a particular part of the program.  Users don't normally follow through with one plan – their plan is always changing according to what is happening around them.

 

Methodology

In her book, Suchman uses an analytic framework to look at the relationship between the user and the machine.  She also uses video technology to record the user and machine interactions.  This, she claims is much better than taking notes.

 

Nardi commented in her book that Suchman reports three ways of conducting their studies,

Suchman and Trigg cataloged their research methods in describing how they conduct empirical studies. What is left out is as interesting as what is included. The authors report that they use (1) a stationary video camera to record behavior and conversation; (2) "shadowing" or following around an individual to study his or her movements; (3) tracing of artifacts and instrumenting of computers to audit usage, and (4) event-based analysis tracking individual tasks at different locations in a given setting (Nardi, 1995).

 

A Case Study

Suchman describes an example of an expert help system which is "a computer-based system attached to a large and relatively complex photocopier, and intended to instruct the user of the copier in its operation.  The study defines the expert help system, the problem of following instructions, communicating instructions, the basic interaction, and the methods. (The following has been taken from Chapter 6 and modified for content.) 

The expert help system 

Before the study begins, some predictions can be made about how the event of using the expert help system will go.  These assumptions can be made because of "certain constraints":

-         The interaction is instrumental;

-         The possible goals of the interaction are defined by the machine's functionality;

-         The structure of the interaction is procedural, constituted by a sequence of actions whose order is partially enforced;

-         The criteria of adequacy for each action can be specified (p. 99).

 

The strategy that the design adopts is to project the course of the user's actions as the enactment of a plan for doing the job, and then use the presumed plan as the relevant context for the action's interpretation (p. 99). Through the user’s response to a series of questions about the state of her original documents and the desired copies, her purposes are identified with a job specification, the specification invokes an associated plan, and the enactment of the plan is prescribed by the system as a step-wise procedure (p. 100).

 

The problem of following instructions 

The practical problem that the expert help system was designed to solve arises out of the work of following instructions, which in turn leads to the work of communicating them (p. 101).  A general observation [from studies of instructions] is that instructions rely upon the recipient's ability to do the implicit work of anchoring descriptions to concrete objects and actions (p. 101).  At the same time, that work remains largely unexamined by either instruction-writer or recipient, particularly when the work goes smoothly (p. 101). More than the correct execution of an instruction, in other words, successful instruction-following is a matter of construction a particular course of action that is accountable to the general description that the instruction provides (p. 102).  The credibility of instructions, moreover, rests on the premise that not only do they describe what action to take, but that if they are followed correctly the action will produce a predictable outcome (p. 102).  An unexpected outcome, accordingly, indicated trouble and warrants some remedy (p. 102).

 

Communicating instructions 

An appreciation for what is required in instruction-following makes it easier to understand the problem that the communication of instructions attempts to solve: namely, the troubles inherent in turning an instruction into an action (p. 104). The restrictions generally associated with written instructions derive not from the writing so much as from the absence of interaction, while the effectiveness of verbal instruction derives less from the speech than from the interaction that is generally associated with it (p. 105).

 

The basic interaction 

The aim of the expert help system is to use the power of the computer to combine the portability of non-interactive instructions, with the timeliness, relevance, and effectiveness of interaction (p. 106).  The machine presents the user with a series of video displays (p. 107).  Each display presented to the user either describes the machine's behavior, or provides the user with some next instructions (p.107).

 

THE BASIC INSTRUCTIONAL SEQUENCE 

 

1. MACHINE PRESENTS INSTRUCTION

    User reads instruction

    interprets referents

    and action descriptions

 

2. USER TAKES ACTION

    Design assumes

    that the action means

    that the user has understood

    the instruction

 

3. MACHINE PRESENTS NEXT INSTRUCTION

 

Methods 

The study was directed by two methodological commitments, one general, the other particular to the problem at hand (p. 109).  Generally, the study began with a commitment to an empirical approach, along with the conviction that situated action cannot be captured empirically through either examples constructed by the researcher, paper and pencil observations, or interview reports (p. 109).  One objective in studying situated action is to consider just those fleeting circumstances that our interpretations of action systematically rely upon, but which our accounts of action routinely ignore (p. 109).  A second objective is to make the relation between interpretations of action and action's circumstances our subject matter (p. 109).

 

On the one hand, the situation was constructed so as to make certain issues observable – specifically, the work of using the machine with the assistance of the expert help system (p. 114).  The construction consisted in the selection of tasks observed to pose problems for new user in "the real world" (p. 114). On the other hand, once given those tasks, the subjects were left entirely on their own (p. 114). The data for this study are corpus of videotapes of first-time users of the expert help system (p. 114). 

 

A second methodological commitment, which arose from the particular problem of looking at human-machine communication, directed the analysis itself (p. 115). The aim of the analysis was to find the sense of “shared understanding” in human-machine communication (p. 115).

 

THE ANALYTIC FRAMEWORK

 

THE USER

THE  MACHINE

Actions not available to

the machine 

Actions available

to the machine 

Effects

available

to the user 

Design rational

 

The framework revealed two initial facts about the relationship of user and system (p. 116).  First, it showed that the coherence of the user’s action was largely unavailable to the system, and something of why that was the case (p. 116).  The richest source of information for the researcher, as a fully fledged, "intelligent" observer, is the verbal protocol (recorded in the first column) (p. 116).  A second, but equally crucial resource is visual access to the user’s actions (p.116).  Of all of her actions, one could clearly see the very small subset, recorded in the second column, that were actually detected by the system (p. 116).

 

The framework proved invaluable for taking seriously the idea that the user and machine were interacting (p. 116).  By treating the center two columns as the mutually available, human-machine “interface”, one could compare and contrast them with the outer columns, as the respective interpretations of the user and the design (p. 116).  This comparison located precisely the points of confusion, as well as the points of true intersection or "shared understanding" (p. 117). 

 

Pros and Cons

The case study above has some interesting observations.  The framework that is used to capture the actions of the machine and the user looks useful.  Instead of just taking notes from the video, the research can use this framework to pull out the specific actions of both the machine and the user as well as what actions were available to the machine and user.  I see this framework as a positive element of situated action.  I don't necessarily like the fact that the user is left alone to figure out the instructions.  I would like the user to be talking aloud about their thoughts and concerns with the instructions and the machine.  This seems very limiting because the researcher doesn't know if the user is confused about the instructions or the machines actions – too much is left for interpretation.  Over all, (from this case study, anyway) situated action can be useful; however, there are some limitations that prevent it from having more potential as a method.

 

It appears that situated action has been accepted as part of HCI; however, it is not accepted as something that should be used alone.  Most agree that it is important to understand the situation of the user and look at the various possibilities that surround them in that particular situation; however that is not enough to design a system.  Some believe that other theories work better to capture the entity of the system and the user.  In an article looking at the neuropsychological side of situated action, Clancey (1993) writes, "SA doesn't require "a whole new language," but it does require that we watch how we use our words, particularly, "memory," "knowledge," "information," "symbol," "representation," and "plan." SA does suggest a different research agenda."  Though an advocate of a new research agenda, Clancey (1993) believes that "it isn't necessary (or perhaps possible) to break completely from traditional theories but instead to reconsider the relation of our models to the cognitive phenomena we sought to understand. 

 

Being an advocate of activity theory, Bonnie Nardi compares situated action to activity theory.  Though she has a few positive comments about the benefits of situated action in HCI, her overall opinion puts activity theory above the rest.

The level of analysis of situated action models—at the moment-by-moment level—would seem to be too low for comparative work.  Brooks criticizes human-factors task analysis as being too low level in that all components in an analysis must "be specified as at atomic a level as possible." This leads to an ad hoc set of tasks relevant only to a particular domain and makes cross-task comparison difficult.  A similar criticism applies to situated action models in which a focus on moment-by-moment actions leads to detailed descriptions of highly particularistic activities (such as pricing cheeses in a bin or measuring out cottage cheese) that are not likely to be replicated across contexts. Most crucially, no tools for pulling out a higher-level description from a set of observations are offered, as they are in activity theory.  Situated action models, then, have two key problems: (1) they do not account very well for observed regularities and durable, stable phenomena that span individual situations, and (2) they ignore the subjective (Nardi, 1995).

In this quote, Nardi recognizes that that a user's situation has to be accounted for, but doesn't believe that it is ad hoc (while promoting activity theory).

Activity theory offers a set of perspectives on human activity and a set of concepts for describing that activity. This, it seems to me, is exactly what HCI research needs as we struggle to understand and describe "context," "situation," "practice." We have recognized that technology use is not a mechanical input-output relation between a person and a machine, that a much richer depiction of the user's situation is needed for design and evaluation. However, it is unclear how to formulate that depiction in a way that is not purely ad hoc. Here is where activity theory helps, by providing orienting concepts and perspectives (Nardi, 1995).

 

Links

More on Situated Action - Short Paper #2 by Ryan Sims

http://minnow.cc.gatech.edu/hci/214

 

Integrating, Not Debating, Situated Action and Computational Models: Taking the Environment Seriously by Michael D. Byrne

http://chil.rice.edu/byrne/Pubs/cogsci94.pdf

 

Even More on Situated Action More on Situated Cognition - Short paper #2 by Lena Mamykina

http://minnow.cc.gatech.edu/hci/219

 

Lucy Suchman's Human/Machine Reconsidered. This paper is a work-in-progress under development as the Introduction to a 2nd, revised edition of Plans and Situated Actions: the problem of human-machine communication. Cambridge University Press, originally published in 1987. http://www.comp.lancs.ac.uk/sociology/soc040ls.html

 

Comments

I can see the place for situated action, but I agree with most of the people I researched and think that situated action should be used with other theories.  Though I see how videos of the situations and using the analytical framework can be helpful, I feel that applying other theories to those situations could only make things better and clarify the situations even further.

 

Most of the research I found was pretty vague about situated action.  They all seemed to have an opinion about it, but didn't really have any substantial evidence or case studies to back up their opinions (neither good or bad).

 

References

Chen, C. & Rada, R. Modelling Situated Actions in Collaborative Hypertext Databases. Retrieved on November 7, 2003, from http://www.ascusc.org/jcmc/vol2/issue3/chen.html

 

Clancey, William J. (1993). Situated action: A neuropsychological interpretation (Response to Vera and Simon). Cognitive Science 17(1):87-107. http://cogprints.ecs.soton.ac.uk/archive/00000459/00/128.htm

 

Doerry, Eckehard. (1995). Evaluating Distributed Environments Based on Communicative Efficacy.  Retrieved on November 7, 2003, from

http://sigchi.org/chi95/Electronic/documnts/doctoral/ed_bdy.htm

 

Nardi, Bonnie. (1995). Context and Consciousness: Activity Theory and Human-computer Interaction. MIT Press.

 

Scharmer, C.O. (1999). Conservation with Lucy Suchman. Dialog on leadership. Retrieved on November 7, 2003, from http://www.dialogonleadership.org/Suchman-1999.html

 

Suchman, L.A. (1987). Plans and Situated Actions: The Problem of Human-Machine Communication. Cambridge: Cambridge Press.

 

Usability First. Retrieved on November 7, 2003, from http://www.usabilityfirst.com/glossary/term_1107.txl