Bruce Edmonds and Contextual Cognition in Social Simulation


How does individual behavior influence the social behavior and vice versa? Social Simulation models (remember eigenbehaviors and Gero’s Situatedness) capture this interplay through various aspects of cognition. (eg social norms) Edmonds argues that in order to better understand this process it is best to contextualize the cognitive processes of the interacting agents. Below the author distinguishes among different types of context:

  • Situational Context: time, location, who was there, the knowledge of the people, the history of the place and all the objects present, “those factors that are relevant to understand this particular occurrence”
  • Linguistic Context: the words that surround an utterance or phrase, elements of the relevant culture, very common as social interactions are composed of linguistic communications. [Peter Gardenfors: Action is primary, pragmatics consists of the rules for linguistic actions, semantics is conventionalised pragmatics and syntax adds markers to help disambiguation (when context does not suffice]
  • Cognitive Context and Framing: some knowledge is acquired in a particular situation and then made available in similar situations and this abstraction is what we refer to as cognitive context, coincides with Goffman’s frames seen as schemata of interpretation used by individuals to locate, perceive, identify and label experiences, the “framing” is a cognitive context of opinion and choice
  • Social Context: events, habits, conventions, norms, a situation that has its own characteristics associated with it.

Learning and Reasoning Edmonds claims, “are far more feasible when their scope is restricted to a particular context“. If, however, one wishes to generalize a thesis, then he/she must be able to appropriately change this focus as the external context, that is the context we inhabit in and thus assess the relevance of knowledge via identifiable “contexts“. This is how we deal with complexity: our limited learning is efficient because many of the possible causes or affects of events that are important remain relatively constant.

The figure above, illustrates what Edmond proposes the architecture of a simulation model should be. There are four modules:

  • Context Identification System (CIS): takes the inputs and learns in a flexible and imprecise way an indication of context
  • Context Dependent Memory (CDM): takes the indication of CIS and identifies all the memory items stored in that context, it evaluates the current truth of these and returns negative feedback to the CIS which will then identify another context
  • Local Learning Algorithm (LL): performs a local update of the knowledge in the memory, propagates successful items towards focus, deletes or corrects items that were false, inserts new items
  • Inference System (IS): it tries to deduce some decisions as to the actions or plans to execute, two common problems, under-determination (not enough info) or over-determination (contradicting info), in the first case the context is expanded, in the second case, the context is reduced.

Edmond’s analysis on the general heuristics are: Formation, Abstraction, Specialization, Content Correction, Content Addition, Content Restriction, Content Expansion, Content Removal

Back to Gero, the heuristics were: Formulation, Synthesis, Analysis, Evaluation, Documentation, Reformulation type 1, Reformulation type 2, Reformulation type 3



  • Edmonds, B., 2014. Contextual Cognition in Social Simulation. In Context in Computing, Patrick Brézillon & Avelino J. Gonzalez (Eds.) New York: Springer DOI 10.1007/978-1-4939-1887-4_18, full paper available here
  • Downes, St., Commentary on Contextual Cognition in Social Simulation, available here

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s