From Jsarmi

Here are some reflections tracing the notion of a "problem space" and how it" can be considered to be a continuous (although contingent) achievement of a group engaged in collaborative problem solving

Contents 1 Problem Space 1.1 Task Environments 1.2 Problem Spaces 1.3 SOAR and Newell's views on the "social band" 1.4 Shortcomings 2 Joint/Shared Problem Space 3 Collaboration and a Dual-problem Sspace 4 Group Cognition 5 Boundaries of a Shared Problem Space 6 Synchronic and Diachronic Problem Spaces 7 References

Problem Space

The idea of a problem space is central to the information-processing or cognitive view of human problem solving (i.e. Newell and Simon's problem-solving theory). From this perspective, the task environment gets abstracted into an internal representation which is what the problem solver uses (i.e. manipulates symbolically) to actually solve the problem. This internal representation is what a subject is assumed to construct when she understands a task correctly and might contain (encode) different forms of knowledge

Task Environments

The information-processing perspective on individual problem solving originated from the collaborative work of Allan Newell and Herbert Simon (Newell & Simon, 1972). In their careful presentation of the theory, Newell and Simon stated their belief that a “process theory” describing the performance of individual “intelligent adults in our own culture,” working on short and “moderately difficult problems of a symbolic nature” (Ibid, p. 3) where “motivation is not a question and emotion is not aroused” (p. 53) could serve as the first step from which an understanding of human thinking could be built up. To achieve this, they explicitly excluded group activity as well as “long-term integrated activities” involving multiple episodes of action over longer periods of time (p.4). Central to the theory is the idea that to solve a task or problem, one must “adapt” to the environment presented by the problem (the “task environment”) by constructing an internal representation of the problem’s relevant elements (a “problem space”), which can be manipulated internally and used to search for a solution. The concept of a “task environment” (sometimes referred as the “external” environment or, simply, the stimulus) is, by definition, an observers’ abstraction. Despite being an analyst’s construct, this abstract environment was conceptualized to be closely shaped by the relevant features of the kinds of subjects who are or will be presented with the task. Newell and Simon conceived the task environment as an objective description of the “vocabulary of elements and relations” that any subject of a particular kind would use or find relevant in his “initial encoding” of the task (p. 58). Relevance, in this sense is relative to the subjects in question. In the case of Newell and Simon’s studies, for instance, the level of competence of the subjects was assumed to be equivalent and treated as not being problematical. This view on relevance, also dictates that certain features of the task have no effect on the way subjects approach it (e.g. the shapes used to identified the marks of the players of a tic-tac-toe game, whether they are X and O or any other symbol) and that such features can be identified and excluded. It is in this sense that the theory views a task as “demanding” certain behaviors from a subject engaged with it. Simon made this concept widely known with his parable of the ant walking on a beach. The complexity and irregularity of the ant’s path is to be attributed to the characteristics of the surface of the beach, not to the complexity within the ant (Simon, 1981). The adequacy of a particular objective description of the task environment is then to be judged by the correspondence between its elements and the responses of subjects engaged with the task, as represented by another abstraction: the problem space.

Problem Spaces

The problem space was introduced as a “neutral and objective way of talking about the responses of the subject, including his internal thinking responses, as he goes about dealing with the stimulus situation” (Newell & Simon, 1972, p. 59). This space, mostly viewed as internal or mental but sometimes related to external resources as well (e.g., Kotovsky & Simon, 1990), is commonly presented as a graph with nodes and links. A subject is assumed to understand a task correctly when he has successfully constructed a problem space representation which contains or encodes: a set of states of knowledge including the initial state of the problem, the goal state, and the necessary intermediate states as well as operators for changing from one state into another, constraints determining allowable states and moves (operators), and any other encodings of knowledge of problem solving methods, heuristics, etc (Ibid, pp. 59 & 810).

• A a set of states of knowledge including:

• the current initial state of the problem, and
• the goal state or condition (as a test procedure or a declaration of what the "end" is) to know when one is done

• Operators for changing one state into another,
• Constraints determining allowable moves (operators) and states (e.g. nodes and allowable links of the space)
• Other representations encoding knowledge of problem solving 'methods, heuristics, or metrics specific to the current task environment, ideas about when one is getting closer to a solution or control knowledge for deciding which operator to apply next, common ways of overcoming stuck states, etc.

Problem solving proceeds as the subject works from the initial state in his evolving mental space, actively creating and exploring possible solution paths, testing and evaluating the results obtained. This process is commonly characterized as “search” on the problem space, and search as an activity becomes the central phenomena theorized. The level of detail in the proposed descriptions of the candidate search processes is, undoubtedly, one aspect in which this theory rivals other less specified proposals. For instance, a number of search methods have been described as being candidate explanations of the processes used by human problem solvers: breadth first, depth first, branch and bound, bidirectional, heuristic best first, hill climbing, etc. Newell further specified that the problem solving process included two separate search processes: problem search and knowledge search (Newell, 1990, p. 98). Problem search, the purposeful creation and traversing of the nodes of the problem space is guided or controlled by knowledge search processes which retrieve previously stored information and, for instance, help in selecting the operators for the problem search and in evaluating their consequences (p. 99). Furthermore, Newell advanced the hypotheses that search in combinatory spaces is the fundamental human process for attainment of difficult tasks and that all deliberate cognitive activity occurs in problem spaces (Newell, 1980). Learning is consider to be a phenomena that happens constantly as humans engage with problem solving (Newell, 1990, p. 312) and involves the acquisition of new operators, new strategies for controlling the search process, declarative data, and even the appropriation of tasks themselves (with their associated problem spaces) mostly seen as being stored and retrieved in memory. The concept of problem space and its associated dynamics have been used to produce detailed descriptions of individual problem solving activity. Newell and Simon’s theory has had a significant and long lasting effect in the way human cognition is characterized. However, applying the concept of problem space to other contexts and types of tasks might require some additional considerations (e.g., Anderson et al., 2000; Kotovsky & Simon, 1990).

SOAR and Newell's views on the "social band"

SOAR, the integrative cognitive architecture derived from Allan Newell's work, achieves all cognition by search in problem spaces, and architecturally supports this by a flexible, two-level recognize-decide-act control structure. Problem spaces are based in part on the idea that search in combinatoric spaces is the fundamental process for attainment of difficult tasks. The nature of such search is seen most easily in tasks like chess that have a well-defined set of operators and states. A search space consists of a set of (generated) representational states and operators that transition between states. Problem spaces as realized in SOAR extend the standard notion of search in an important direction: problems spaces are taken to be the fundamental way that humans accomplish all cognitive tasks, including routine (i.e, well-practiced) tasks. SOAR is therefore one realization of the problem space hypothesis (Newell, 1980b), which asserts that all deliberate cognitive activity occurs in problem spaces.

According to Newell:

As the time scale increases from days to weeks to months, the systems involved become social. They comprise multiple individuals in interaction. Humans do not lead solitary lives at the level of days and above. Thus, the system just above the rational level is the social face-to-face group and it moves up to encompass more distal organizations as the time scales increases further. The issue of continued upward extension and perfection of rationality is whether groups can be described as knowledge-level systems —as a single body of knowledge with a set of goals— such that the group’s behavior is explainable and predictable by its use of knowledge in the service of its goals. In addition, the approximation should become better as more time becomes available.

That rationality fails for groups we all know directly from our own experience. The kernel of the reason appears to be the small bandwidth of communication between human beings. The components of a social system —the humans— have too much knowledge relative to how rapidly they can communicate it to each other (The constraint on communication indicates one difficulty with the use of society as a model for mind and intelligence (Minsky, 1986). The society-of-minds metaphor is built around relatively small communication channels, but inside each mind it is not that way at all). There is no way for a social group to assemble all the information relevant to a given goal, much less integrate it. There is no way for a social group to act as a single body of knowledge. This restriction applies to goals as well. Whatever advantages may be found in reconciling opposed or discordant goals, the limited means of communicating about goals and why they exist is sufficient to keep the organization from ever arriving at a single common basic goal structure.

(…)

The net effect is that the social band becomes characterized as a distributed set of intendedly rational agents, in which each agent has a large body of knowledge relative to how fast it can communicate it –that is, there is limited information flow. This characterization hardly accounts for everything in the social band, but it does account for a number of things. P. 153-154.

Shortcomings

This conception of problem-solving as search through a problem space "localized" in the head of an individual was specially insufficient when analyzing problem-solving in everyday life. In part, this difficulty arises from the fact that the information processing model was constructed from a "knowledge lean" approach in which most of the procedures used are context-free. In addition, the applicability of the model seemed to be less straightforward in cases where problems were not as well structured. For example, when Lave studied mathematical activity in supermarkets she observed how such activity was centrally rooted in "being “in” the “store”, walking up and down “aisles”, looking at “shelves” full of cans, bottles, packages and jars of food, and other commodities." Many other similar investigations challenged the restrictive notion of what should be taking into consideration when speaking of a "problem space." Some of these challenges also came from investigations of collaborative problem-solving and the use of artifacts during problem-solving activity.

SHORT TERM, LONG TERM MEMORY, AND MOVEMENT IN MEMORY. If problem solving is search, we should say that search depends on "fetching" or "moving" (or retrieving) the right pieces of information and purposefully storing them in the right place.

Many situations that are recognized as "problem solving" activity by the participants engaged in them, have no recognizable fixed set of operators that could define a unique task environment or problem space, and the constraints, evaluation criteria, and valid states are usually worked out through local interactions. Interestingly, when participants engaged in these kinds of activities they are commonly recognized as doing "learning work" or "knowledge building." Understanding such situations, the resources and activities that seem most prominent in them, offers a crucial opportunity to specify the dynamics of situated problem solving and knowledge building. As some have critically pointed out, perhaps even the most notable attempts at describing such activites still "owe us the details" (Kirsh, in press, p. 35) Planning and execution are complex inter-related activities that might constitute each other as sense making is conducted (SUCHMAN, GREENO)

HOW DID THE GROUPS FRAME PROBLEMS FROM THE SITUATION? Some started from a familiar question (distance) in some cases ignoring some of the constraints and evolved their inquiry by refining it, adding constraints (shortest through the staircase), adding features (3D). Call this "Interactive Registration". ADDING STRUCTURE TO THE ENVIRONMENT. Call this creating a "feature rich situation" Call this indexical entrailment. EXAMPLE: Counting of the paths. What happens here? The grid is a graphical resource that is available, visually to all. The contraints on how manipulation of the grid is to be conducted are, in part, describe in the problem statement. Explorations and manipulations can happen by tracing paths on the grid and most likely happen, as an interactive and purposeful engagement of what is perceived from the figure and the reasoning procedures active at the moment. (OBSCURE?) In the 2 by 2 there are 6, I think? I can only see four. HOW DO WE EXPLAIN THIS? The subject had 6 in Long term memory, retrieved that fact to Short Term memory but then when looking at the figure, it couldn't test this fact so had a mismatch that motivated a breakdown. Subject B, sees 6. What does that mean? Possibly, that he has deploy a reasoning procedure to find, diferentiate and count ways of navigating the paths between a and b. What is in the problem space? NOW AND NOW? Where is the problem space? NOW? AND NOW? Explore this sequence -> A has an incomplete PSpace because he only sees 4. B has a complete one because he sees 6. B starts to show A, and that "showing" is in the external world, but before he finishes A has it.. has constructed it. There are two bystanders, where are their problem spaces? Somebody else can see this later... where is their problem space? What does it contain? IS it joint? Kirsch distinguishes between PRoblem space and World... or maybe problematizes this distincitio? OR at least challenges us to problematize it and provide better descriptions of how participants orient to that distinction and how the methods they used are consequential in shaping their experience. And then there is the "social" or "relational" but Where as Teasley and Roschelle seemed to imply that the relational dimmension was used as a resource to advanced the "content" Barron seems to imply that sometimes they compete for attention, or that one can be detrimental to the other.

What counts as part of the problem space in situated cognition? The problem is that "all context" is a big and amorphous container and we don't know where to start. But we can pay attention to what the participants orient to, especially in collabirative interacitons! So propose that approach in order to better specify what "ought" to be included in the problem space. PRoblem space as an interacional achievement. And how does problem space accounts for lerarning?????

The position from situated HOW DOES THEIR ATTENTION TO TIME AND SEQUENCE INFORMS OUR DECISION OF WHAT IS IN THE PROBLEM SPACE?

Joint/Shared Problem Space

Roschelle and Teasley proposed the notion of a joint problem space (JPS) as a "shared knowledge structure" that supports problem solving activity in collaborative contexts by integrating:

• Goals
• Descriptions of the current problem state
• Awareness of available problem solving actions

Initially, this description does not seem too radically different than the traditional cognitive view. After all, one can conceive each one of these elements as internal representations and the "shared knowledge structure" as the intersection of these individual "mental models" or individual problem spaces. In fact, this has been the predominant psychological interpretation of collective cognition (e.g. Fiore and Salas' theory of team cognition'). However, Roschelle and Teasley's proposal goes beyond this simple interpretation. From their perspective "social interactions in the context of problem-solving activity occur in relation to a "shared conception of the problem" which results from a collaborative process of coordination of communication, action, and representation in a particular context of activity, not in the heads of the individuals. This perspective, and indeed, their method of analysis, are consistent with Garfinkel's ethonmethodological observation about the nature of "shared agreements:"

"Shared agreement" refers to various social methods for accomplishing the member's recognition that something was said-according-to-a-rule and not the demonstrable matching of substantive matters. The appropriate message of a common understanding is therefore an operation rather than a common intersection of overlapping sets. (p.30)

From this perspective, a "shared agreement" or a "mutual conception of the problem" is then the emergent and situated result of the participant's interactions, embodied in their context of activity. In the words of Roschelle and Teasley, it is "the coordinated production of talk and action by two participants (what) enabled this construction and maintenance (of the joint problem space) to succeed... the introduction of successful ideas was sometimes asymmetric, although it succeeded only through coordinated action." One analytical problem resulting from this perspective is that of situating the three elements of the joint problem space (goals, current problem state, and awareness of available actions) in interaction and linking such interactional products with the cognitive reality of individuals.

Collaboration and a Dual-problem Sspace

More recently, other researchers have furthered the notion of a joint problem space adding more evidence to the importance for social interaction in the construction, use and maintenance of such resources. Baron (2003), for instance, argued beyond the need to "manage joint attention at solution-critical moments" as explored by Roschelle, groups face other complex challenges that when successfully managed accounted for their overall advantage over groups with equivalent capacity. In particular, she argued that there are critical challenges to the creation of joint problem-solving spaces as individual participants are simultaneously managing their own efforts to understand pieces of the problem and at the same time trying to understand what others are doing. As a result, Baron argues that "collaboration might productively be thought of as involving a dual-problem space that participants must simultaneously attend to and develop." This dual space" comprises:

• A content space consisting of the problem to be solved, and
• A relational space consisting of the interactional challenges and opportunities.

Baron argues that the content space and relational space are negotiated simultaneously but can compete for limited attention, "one needs to be able to monitor and evaluate one’s own epistemic process while tracking and evaluating others’ epistemic processes." An example of how these two spaces intersect and one that Baron follows through her entire analysis of group problem solving is the management of problem-solving proposals. A problem-solving proposal has both a relationship with the content space or problem space and to the social interaction of the group. It involves elements of the problem or of a strategy that might require attention, offer a new insight, or might indicate a solution path. At the same time, they are "bids" for collective action, requests for co-participants to, perhaps, interrupt their current activity, and in some cases, contrasting alternatives to other proposals already stated. This "dual" character of the interaction space expands our understanding of what a "joint problem space" entails, although, it still remains as a challenge to study their interdependence in-depth.

Group Cognition

The dialectical relationship between social interaction and the construction of meaning (i.e. a problem space) has been directly addressed by the theory of group cognition (Stahl, 2006a). From this perspective, the organization of action is an emergent property of moment-by-moment interactions among actors, and between actors and the activity system in which they participate collectively. The content space and the relational space are mutually constitutive from this perspective. Stahl, offers a candidate description for how the dynamic process of building knowledge might intertwine the content and relational spaces:

"Small groups are the engines of knowledge building. The knowing that groups build up in manifold forms is what becomes internalized by their members as individual learning and externalized in their communities as certifiable knowledge. At least, that is a central premise of this book." (Stahl, 2006, p. 16)

Stahl, in his analysis of how groups "sustain" their group cognition while engaged in online mathematical problem solving, has alluded to two dimensions of how time might be an important element of problem-solving activity. On the one hand, the collaborative activity involved in solving a problem can be "spread across" hundreds of micro-level interactions. On the other hand, individuals might internalize the meaning co-constructed through interactions and "sustain" the group cognition by engaging in later individual or group work. In either case, "groups sustain their () social and intellectual work by building longer sequences of math proposals, other adjecency pairs and a variety of interaction methods."

Interaction, here is taken in the full sense that ethnomethodologists give it, as the "ongoing, contingent co-production of a shared social/material world," and which, as Suchman argues "cannot be stipulated in advance, but requires an autobiography, a presence, and a projected future (Suchman, 2003).

From this perspective, different aspects of the reality of groups solving problems are illuminated. The following concepts attempt to outline processes and resources that may be used to described collaborative problem-solving as human interaction:

• The problem-as-presented and the problem-to-be-solved. Initially, a problem might be a set of textual, graphical and other resources available for a group to consider. These resources are used in interaction, and through interaction become a problem-to-be-solved involving a group's own "problematization" (Koschmann et al, 2003) of individual parts of the problem-as-presented or its entirety. This activity can be re-visited multiple times during the interaction with the problem-to-be-solved evolving or changing. This is what in traditional theory of problem-solving was referred as "framing" but was rarely studied in any direct way.

• Proposals and courses-of-action. Problematization defines a focus of attention and establishes an initial set of relevancies but it is through exploration, the development of proposals and the interactive engagement with them that problem-solving evolves. Describe here how proposals can become shared courses of action contributing to the development of meaning-objects (e.g. math objects) and collectivities. Explore how proposal work monitors the proposed activity and the actions-according-to-plan ala Suchman

• Meaning-objects. math objects, observations, constraints, etc. discovered and stated in interaction to shape the context of activity

• Solution objects. objects constructed and treated as solutions whether they actually "solve" the problem in an abstract sense or not

• Collectivities-in-action Alignments, orientations, participation frameworks, ephemeral division of labor, etc.

• Trajectories-of-action how to start or end something, Awareness of where the group has been, where it could go next, where others have been, etc.

Boundaries of a Shared Problem Space

The distinction between the cognitive view of a problem space as an idealized structure in someones head and a problem space as the emergent of people engaged in problem solving activity, can be explored from the perspective of boundaries:

• Where does the problem space reside?
• When does the problem space can be said to no longer exist?

(Note: Distributed Cognition may have a view of a distributed problem space that is also relevant here?)

Synchronic and Diachronic Problem Spaces

Here is a possible distinction of the work of groups, pased on both their synchronous problem space and their diachronic problem space

References

Barron, B. (2003). When Smart Groups Fail. Journal of Learning Sciences 12(3), 307–359.

Fiore, S. M., & Salas, E. (2004). Why we need team cognition. In E. Salas & S. M. Fiore (Eds.), Team Cognition: Understanding the Factors that Drive Process and Performance, (p. 235-248). Washington, DC: American Psychological Association.

Garfinkel, H. (1967) Studies in Ethnomethodology. Englewood Cliffs, NJ: Prentice- Hall. (Some essential features of common understandings)

Greeno, J. (2006). Learning in Activity. In R. K. Sawyer (Ed.), The Cambridge Handbook of the Learning Sciences (pp. 79-96). Cambridge: Cambridge University Press.

Koschmann, T., Zemel, A., Conlee-Stevens, M., Young, N., Robbs, J., & Barnhart, A. (2003). Problematizing the problem: A single case analysis in a dPBL meeting. In B. Wasson, S. Ludvigsen & U. Hoppe (Eds.), Designing for change in networked learning environments: Proceedings of the international conference on computer support for collaborative learning (CSCL '03) (pp. 37--46). Amsterdam: Kluwer Academic Press.

Lave, J. (1988). Cognition in Practice: Mind, Mathematics and Culture in Everyday Life. Cambridge University Press, Cambridge, UK.

Newell, A. (1980). Reasoning, problem solving and decision processes: The problem space as a fundamental category. In R. Nickerson (Ed.), Attention and Performance VIII . Hillsdale, NJ: Erlbaum.

Newell, A. (1990). Unified Theories of Cognition. Cambridge, Massachusetts: Harvard University Press.

Roschelle, J. & Teasley, S.D. (1995). Construction of shared knowledge in collaborative problem solving. In C. O’Malley (Ed.), Computer-supported collaborative learning. New York: Springer-Verlag.

Stahl, G. (2006a). Group cognition: Computer support for building collaborative knowledge. Cambridge, MA: MIT Press.

Stahl, G. (2006b). Sustaining Group Cognition in a Math Chat Environment. Research & Practice in Technology Enhanced Learning, 1(2), 85-113.

Suchman, L. (2003). Reading and Writing: A response to comments on Plans and Situated Actions. The Journal of the Learning Sciences, Vol. 12, No. 2: 299-306.