ACT-R Cognitive Architecture

actI came across ACT-R and determined it to be worthy of a look when two arguing European psychologists spoke of it approvingly.  What is it?  The basis of the acronym is Adaptive Control of Thought – Rational. It was created by John Anderson of Carnegie Mellon.  Carnegie Mellon is known by me to be one of if not the preeminent computer science school and ACT-R has roots in artificial intelligence.

According to its website, ACT-R is a cognitive architecture: a theory about how human cognition works. On the exterior, ACT-R looks like a programming language; however, its constructs reflect assumptions about human cognition. These assumptions are based on numerous facts derived from psychology experiments.

Like a programming language, ACT-R is a framework: for different tasks, researchers create models (aka programs) that are written in ACT-R and that, beside incorporating the ACT-R’s view of cognition, add their own assumptions about the particular task. These assumptions can be tested by comparing the results of the model with the results of people doing the same tasks. The “results” are the traditional measures of cognitive psychology:

  • time to perform the task,
  • accuracy in the task, and,
  • (more recently) neurological data such as those obtained from fMRI.

One important feature of ACT-R that distinguishes it from other theories in the field is that it allows researchers to collect quantitative measures that can be directly compared with the quantitative measures obtained from human participants.

ACT-R has been used successfully to create models in domains such as:

  • learning and memory, especially known for learning algebra and geometry.
  • problem solving and decision making, including driving while talking on a cell phone and air traffic control.
  • language and communication,
  • perception and attention,
  • cognitive development, or
  • individual differences.

One critical point is that in using ACT-R, one is forced to make the theory computationally explicit, thus allowing for true evaluation of the theory as well as allowing for predictions in novel circumstances.

According to the ACT-R theory, cognition emerges through the interaction of several relatively independent modules. Figure 1 illustrates the modules in a model of solving equations such as 7x + 3 = 38. In Figure 1, the visual module is extracting information about the equation; the retrieval module is obtaining arithmetic facts relevant to this information; and the imaginal module is changing the representation of the solution to incorporate the retrieved information.

When interpreting this research, it is important to understand the basic processing cycle that involves all of the modules associated with these regions in the ACT-R
theory. At any point in time the state of the system is defined by the contents of the buffers of these modules. Because of the way computations are distributed across
modules, four basic operations tend to repeat when moving to a new state:
(i) Procedural: some mental action is selected (e.g. add a number to both sides of an equation) that is appropriate to this current state.
(ii) Goal: this might result in some change in the goal that is controlling the current step (e.g. get rid of the + 3 before the x).
(iii) Retrieval: frequently, it will be necessary to retrieve information (e.g. an addition fact) from declarative memory.
(iv) Imaginal: the problem representation is often updated to incorporate the retrieved information (e.g. ‘7x = 38 – 5’ to ‘7x = 35’)

Figure 1. A representation of the basic module operations in ACT-R to implement the unwind strategy in ACT-R to solve the equation 7*x + 3 = 38. (i) The visual module encodes pieces of the visual display such as fragments of an equation, for example ‘+ 3’. (ii) The retrieval module holds retrieval cues such as ‘8 – 3’ to drive the retrieval of task-relevant facts. (iii) The imaginal module creates and transforms problem representations, such as intermediate answers in the equation solution. (iv) The goal module sets control states to direct the path of information processing, such as whether information is to be retrieved or the equation is to be transformed. (v) The manual module programs the output such as the keying of 5 as the final answer. (vi) The procedural module executes productions that recognize patterns of activity in other modules, selects appropriate actions and relays information to the other modules. The height of the boxes in Figure 1 represents the time a module is active in doing things such as retrieval or constructing an internal problem representation. While a module is engaged during one of these activities, it might be performing a great many computations in
parallel to achieve its objectives, such as the retrieval module matching a pattern against declarative memory. It places the results of its computation in its buffer associated for access by other modules. Multiple modules can work in parallel, but the need to pass information among modules imposes some seriality on the overall processing.  This can also create bottlenecks.

ACT-R describes cognition as a set of modules that interact through a production system. The production system consists of production rules (i.e., if-then rules) whose conditions (i.e., the “if” parts) are matched against the contents of the modules. If a rule’s conditions are met, then the production rule can fire and the specified action is carried out. In coordinating the modules, the production rules can only act on information that is available in buffers, which can be thought of as processing bottlenecks, linking the modules’ contents to the production rules. For instance, the production rules cannot access all contents of the declarative module, but only these that are currently available in the retrieval buffer.

Julian Marewski et al (see below) explained how they used ACT-R in their exploration of cognitive niches. They noted that all models perform the same task as their experimental subjects: The models “read” two city names off a computer screen, process them, decide for one of them, and enter a response by “pressing” a key. The various decisional, memorial, perceptual and motor processes assumed by the model are coordinated by production rules. Specifically, by first “reading” the names of both cities, the model tries to retrieve a memory trace of the city names called a chunk. Chunks are facts like “York is a city” or “York has an airport” which model people’s familiarity with city names and their cue knowledge about these cities, respectively. For each cue, the model retrieves its validity. If the cue value is positive, the model adds the validity of this cue to the weighted sum of the city, initiating a summation procedure. If the cue value is negative, the model subtracts the validity of the corresponding cue from the weighted sum of that city, initiating a subtraction procedure. Finally, the model compares the total weighted sums of the two cities and chooses the one with the larger total weighted sum by pressing a key.  Marewski and Andreas Glockner were the arguing psychologists who pointed me toward learning a little about ACT-R.  In a future post, I intend to discuss their arguments.

So the best that I can tell, ACT-R is a framework based on certain ideas of how the brain works that lets researchers test their models.  It seems to impose structure on models by making them explicit.  ACT-R remains a work in progress, but one with some legitimacy at least as described by Anderson (see below). Anderson writes that although the original ACT-R predated extensive use of fMRI, the imaging data have had a major influence on the ACT-R theory at two levels. At the level of the architecture, they have helped  to articulate the current modular structure of ACT-R. For instance, imaging data helped indicate the need for a distinction between the goal and imaginal modules, which had been conflated into a single system in earlier versions of ACT-R. At the level of specific models within the architecture, imaging data have helped guide modeling decisions.

“A central circuit of the mind”, John R. Anderson, Jon M. Fincham, Yulin Qin and Andrea Stocco in Trends in Cognitive Science, Vol 12, No. 4, pages 136-143.

“Constraining ACT-R Models of Decision Strategies: An Experimental Paradigm”
Cvetomir M. Dimov, Julian N. Marewski, and Lael J. Schooler, as presented at COGSCI 2013.