MODELING HUMAN PERFORMANCE

Anderson et al. Article and Taatgen, Lebeire, & Anderson Article

Presented by : Kelsey Baldree

INTRODUCTION

In order to specialize, viewing the mind as consisting of a set of specialized components Why is this view point important to the field of human factors and psychology?

Newell’s argument for an integrated system A single system (mind) produces all aspects of behavior. It is one mind that minds

them all. Even if the mind has parts, modules, components, or whatever, they all mesh together to produce behavior. Any bit of behavior has causal tendrils that extend back through large parts of the total cognitive system before grounding in the environ-mental situation of some earlier times. If a theory covers only one part or component, it flirts with trouble from the start. It goes without saying that there are dissociations, ndependencies, impenetrabilities, and modularities. These all help to break the web of each bit of behavior being shaped by an unlimited set of antecedents. So they are important to understand and help to make that theory simple enough to use. But they don’t remove the necessity of a theory that provides the total picture and explains the role of the parts and why they exist (pp. 17-18)

INTRODUCTION

There are 2 advantages for the unified field theory1. Concerned with producing a theory that is capable of attacking real-

world problems2. Concerned with producing a theory that is capable of integrating the

mass data from cognitive neuroscience methods Why are these advantages important for developing a theory for an

integrated theory? Central to ACT-R is the notion of a declarative memory for facts, and a

procedural memory for rules

ACT-R 5.0 ARCHITECTURE

ACT-R consists of a set of modules, each of which are devoted to processing a different kind of information Some of the modules in ACT-R are

similar to the modules in the system Coordination in the behavior of these

modules is achieved through a central production system Which is not sensitive to most of the

activity, but rather can only respond to a limited amount of information Why is it important the central system perform

like this? Core production system can recognize

patterns in these buffers and make changes to the buffers

ACT-R 5.0 ARCHITECTURE

The theory is not committed to exactly how many modules there are, but rather a number have been implemented as part of the core system Is this good to leave the theory open-ended?

The buffers of the modules hold limited information that the production system can respond to Why would there only be limited information held? ACT-R 5.0 includes a theory of how these buffers interact to determine

cognition There are at least 5 regions of the frontal cortex which play a major role in

controlling behavior The critical cycle in ACT-R is one in which the buffers hold representations

determined by the external world and internal modules The assumption in ACT-R is that a cycle takes about 50 ms to complete

ACT-R 5.0 ARCHITECTURE

Conditions of the production rule specify a pattern of activity in the buffers that the rule match and the action specifies changes to be made to the buffers In order for this to happen, the architecture assumes there is a mixture of parallel and serial processing There are 2 levels of serial bottlenecks

1. Chunks2. Only a single production is selected at each cycle to fire

Although each is its own separate system, each contributes to the overall integration of cognition

THE PERCEPTION-MOTOR SYSTEM

Although not focused on perception or action, the division of labor of the system tends to lead to a treatment of cognition that is abstracted from the perceptual-motor system Why would this be?

ACT-R adapted the strategy used for model human processor system defined by Card, Moran, and Newell (1983) and reimplemented certain aspects of the EPIC system Although the strategy will breakdown at points, the system has proven

to be quite workable Why would this be? Is this the best way to construct a module?

THE GOALS MODULE

Although human cognition is embodied, its embodiment is not what gives human cognition its advantage over other species. Its advantage depends on its ability to achieve abstraction in content and control Research indicates the prefrontal regions pay an important role in maintaining the goal state In a follow up (Fincham et al., 2002) found the DLPFC, bilateral

parietal regions, and the premotor cortex show response to a number of planning subgoals Supports the idea that goal functions are maintained across multiple brain

regions If the goal functions are distributed, would the hypothesis of a single goal

structure still hold true?

THE DECLARATIVE MEMORY MODULE

Declarative memory promotes long-term personal and cultural coherence

Access to information in declarative memory is hardly instantaneous or unproblematic, declarative memory is an important component of the ACT-R theory which concerns the activation processes that control this access The declarative memory system which constitutes the cognitive core of ACT-R can have their behavior controlled by a set of equations and parameters

ACTIVATION EQUATION

The sum of a base-level activation reflecting its general usefulness in the past, and an associative activation reflects its relevance to the current content Activation chunk : A Base-level activation : B Attentional weightings : W Strength of association : S

Activation chunk controls both its probability of being retrieved and its speed of retrieval

BASE-LEVEL ACTIVATION EQUATION

The base-level activation raises and falls with practice and decay Time since jth practice : t

Based on rational analysis of Anderson and Schooler (1991)

Reflects the log odds an item will reoccur aa a function of how it has appeared in the past

MAPPING ACTIVATION EQUATIONS

There are 2 equations which map activation1. Probability of retrieval

Assumption is chunks will be retrieved only if their activation is over threshold Where s controls the noise in activation

levels Typically set at about 0.4

2. Latency of Retrieval If chunk is successfully retrieved, the

latency of the retrieval will reflect the activation

Latency retrieval at threshold is approximately 0.35 s

PROCEDURAL MEMORY

Act-R consists of a set of modules that progress independently of one another This would be a fragmented concept except for what?

The description in this section focuses on how the production system achieves control and how its adaptive Key point is at any point in time multiple production rules may apply, but because of the seriality in production rule execution, only on can be selected, and this is the one with the highest utility Why do you think its important to choose the rule with the highest

utility?

CALCULATIONS FOR UTILITY

Production Utility Equation P : estimate of probability that if

production I is chosen the current goal will be achieved

G : the value of that current goal C : estimate of the cost to

achieve goal

Production Choice Equation The summation is over all

applicable production and t controls the noise in utilities The value of t is about 0.5

COST EQUATION

The value of the cost parameter is estimated by the sums of the efforts invested in a goal divided by the total number of experiences. Is there a simpler way to put

understand all the equations? Learning mechanisms adjust the

cost and probabilities that underline utilities.

UTILITY LEARNING

A useful mechanism in tasks where there are multiple cognitive strategies, but it is unclear which one is the best strategy.Basic set up is to have a set of production rules for each of the strategies One of these production rules initiates the strategy, which competes with rules which initiate other strategies Does this sound like a big loop to anyone else? But as the rules gain experience, the parameters for the rules will reflect

their utility

Making the system overall sensitive to changes in utility of strategies Is this a positive or negative addition to the model?

EXAMPLES

Anderson et al. article The Effects of Instructional and Practice in a Dynamic Task

The Anti-Air Warfare Coordinator Tracking Multiple Buffers in an fMRI Study

Taatgen, Lebiere, and Anderson article Building Sticks Task Sugar Factory

Which example helped you understand the model the best?

CONCLUSIONS

The concern for the ACT-R architecture is it is an illustration of the potential of integration architectures rather than a final answer

ACT-R’s research focus and most modeling projects involve the development of cognitive models that produce predictions that are matched to human data ACT-R can be used to program agents which exhibit human-like behavior or serve as a theoretical basis to allow agents to construct a model of their users. Is being able to match to human-like behavior create a good enough

model for us to understand our users? All aspects of human cognition are important in producing human-like agents