John Anderson was born in 1947 in Vancouver, British Columbia. He earned his PhD from Stanford University in 1972 and has been a professor at several universities, including Carnegie Mellon University since 1978. ACT* is a cognitive theory developed by Anderson that describes a spreading activation model of semantic memory combined with a production system for executing higher-level operations. It distinguishes three types of memory - declarative, procedural, and working memory - and three types of learning.

1. Biography
John Anderson was born in Vancouver, British Columbia, in 1947. He entered the University of
British Columbia with hopes to become a writer, but left with the dream of practicing psychology as a
precise and quantitative science. He graduated at the head of his class in Arts and Science in 1968.
Anderson earned his Ph.D. from Stanford in 1972 under Gordon Bower. He then spent one
year at Yale as an assistant professor, three years at the University of Michigan as a Junior Fellow, one
year at Yale as an associate professor, and a final year as a full professor. He has been at Carnegie
Mellon University since 1978. ACT* (Pronounced "A-C-T-star") is a cognitive theory dealing primarily with
memory structures.
Theory
The model describes a spreading activation model of semantic memory, combined with a
production system for executing higher level operations. According to this theory, there are three types of
memory and three types of learning.
Declarative memory (WHAT) encompasses factual components and their associations and
sequences.
Procedural memory, or production memory, (HOW) are sequences of behaviors (productions)
based on conditions and actions stored in declarative memory. A production is a series of "if - then" rules: if
x happens, then do y. New productions are formed by linking up existing ones, adding components, and
deleting components.
Working memory is the part of long-term memory which is currently in consciousness. These
three aspects of memory work closely together, and each has its own functions and processes.
act
Three types of learning are
Generalization- in which procedures (productions) are cross-contextualized or more widely
applied
Discrimination - in which procedures (productions) become more specialized
Strengthening - in which procedures (productions) are applied more frequently.
The theory includes notions of goal structure, problem-solving context, and feedback.
Research with ACT* has showed that reaction time for fact retrieval increase as a function of the number of
times the items sought were mentioned in a story. Unique content in stories is easier for the reader to
retrieve.
Memory ACTIVATION determines the probability of access to memory, and the rate at which a
memory can be accessed, after a subject is cued to recall information. Two factors influence the level of
activation: how recently the person has accessed the memory, and how much they have practiced or
rehearsed the information.
SPREADING ACTIVATION proposes that activation travels along a network of connections, so
that once cued, a subject may have multiple responses based on the connections among bits of
information in memory. Spreading activation is not believed to be entirely under the subject's control, but
cueing may activate remote connections without the subject's volition being involved. This tendency for
memories to be activated is called ASSOCIATIVE PRIMING.
ACT-R is a general theory of cognition developed by John Anderson and colleagues at
Carnegie Mellon Univeristy that focuses on memory processes . It is an elaboration of the original ACT
theory (Anderson, 1976) and builds upon HAM, a model of semantic memory proposed by Anderson &
Bower (1973). Anderson (1983) provides a complete description of ACT-R. In addition, Anderson (1990)
provides his own critique of ACT-R and Anderson (1993) provides the outline for a broader development of
the theory. See the CMU ACT site for the most up-to-date information on the theory.
ACT-R distinguishes among three types of memory structures: declarative, procedural and working
memory. Declarative memory takes the form of a semantic net linking propositions, images, and
sequences by associations. Procedural memory (also long-term) represents information in the form of
productions; each production has a set of conditions and actions based in declarative memory. The nodes
of long-term memory all have some degree of activation and working memory is that part of long-term
memory that is most highly activated.
According to ACT-R, all knowledge begins as declarative information; procedural knowledge is learned by
making inferences from already existing factual knowledge. ACT-R supports three fundamental types of
learning: generalization, in which productions become broader in their range of application, discrimination,
in which productions become narrow in their range of application, and strengthening, in which some
productions are applied more often. New productions are formed by the conjunction or disjunction of
existing productions.
Application
ACT-R can explain a wide variety of memory effects as well as account for higher order skills such as
geometry proofs, programming and language learning (see Anderson, 1983; 1990). ACT-R has been the
basis for intelligent tutors (Anderson, Boyle, Farrell &Reiser, 1987; Ritter et al, 2007).
Example
One of the strengths of ACT is that it includes both proposition and procedural representation of knowledge
as well as accounting for the use of goals and plans. For example, here is a production rule that could be
used to convert declarative sentences into a question:
IF the goal is to question whether the proposition (LVrelationLVagentLVobject) is true THEN set as
subgoals
1. to plan the communication (LVrelationLVagentLVobject)
2. to move the first word in the description of LVrelation to the beginning of the sentence
3. to execute the plan
This production rule could be used to convert the sentence: "The lawyer is buying the car." into the
question: "Is the lawyer buying the car?"
Principles
1.
Identify the goal structure of the problem space.
2.
Provide instruction in the context of problem-solving.
3.
Provide immediate feedback on errors.
4.
Minimize working memory load.
5.
Adjust the "grain size" of instruction with learning to account for the knowledge
compilation process.
6.
Enable the student to approach the target skill by successive approximation.
The Adaptive Character of Thought - Rational (ACT-R) is a theory of cognition developed
principally by John Anderson at Carnegie-Mellon University [4]. ACT-R models how humans recall
chunks" of information from memory and how they solve problems by breaking them down into subgoals
and applying knowledge from working memory as needed.
1.1 Roadmap
We rst introduce the crucial distinction between declarative knowledge and procedural
knowledge in Section2. The document then proceeds in a top-down fashion: Under the assumption that the
agent (a human,or possibly a computer) already has all of the knowledge he/she needs, we examine in
Section 3 how the decision-making process is made on a rational basis under ACT-R. In particular, we
describe the mechanism by which a particular production rule," corresponding to the actions" of ACT-R, is
chosen out of many possible alternatives. In Section 4, we remove the assumption that knowledge is
already available, and describe the ACT-R processes by which new knowledge is acquired. This includes
both the creation of new memories, as well as the strengthening (and decay) of existing ones. Finally, in
Sections 6 and 7, we discuss how ACT-R partially models the Spacing Effect and the Power Laws of
Learning/Forgetting.
12 Declarative versus Procedural Knowledge
Under ACT-R, human knowledge is divided into two disjoint but related sets of knowledge {
declarative and procedural. Declarative knowledge comprises many knowledge chunks, which are the
current set of facts that are known and goals that are active. Two examples of chunks are The bank is
closed on Sundays," and The current goal is to run up a hill." Notice that each chunk may refer to other
chunks. For instance, our first example chunk refers to the concepts of bank," closed," and Sunday,"
which presumably arethemselves all chunks in their own right. When a chunk i refers to, or is referred to
by, another chunk j, then chunk i is said to be connected to chunk j. This relationship is not clearly defined
in ACT-R { for instance, whether the relationship is always symmetrical, or whether it can be re
exive (i.e., a chunk referring to itselfin recursive fashion), is not speci ed.
Procedural knowledge is the set of production rules { if/then statements that specify how a
particular goal can be achieved when a specied pre-condition is met { that the agent currently knows. A
productionrule might state, for instance, If I am hungry, then eat." For the domain of intelligent tutoring
systems,for which ACT* and ACT-R were partly conceived, a more typical rule might be, If the goal is to
prove triangle similarity, then prove that any two pairs of angles are congruent." The human memory
contains many declarative knowledge chunks and production rules. At any point in time, when a person is
trying to complete some task, a production rule, indicating the next step to take in order to solve the
problem, may re" if the rule's pre-condition, which is a conjunction of logicalpropositions that must hold
true according to the current state of declarative memory, is fulfilled. Sincethe currently available set of
knowledge chunks may fulfill the pre-conditions of multiple production rules,a competition exists among
production rules to select the one that will actually fire. (This competition willbe described later.)
Whichever rule ends up firing may result either in the goal being achieved, or in thecreation of new
knowledge chunks in working memory, which may then trigger more production rules, andso on.
Process columns" is pushed onto the goal stack."
ACT* THEORY
Explaining memory effects
History and Orientation
ACT* is a general theory of cognition developed by John Anderson that focuses on memory processes.
ACT* distinguishes among three types of memory structures: declarative, procedural and working memory.
Declarative memory takes the form of a semantic net linking propositions, images, and sequences by
associations. Procedural memory (also long-term) represents information in the form of productions; each
production has a set of conditions and actions based in declarative memory. The nodes of long-term
memory all have some degree of activation and working memory is that part of long-term memory that is
most highly activated.
Core Assumptions and Statements
According to ACT*, all knowledge begins as declarative information; procedural knowledge is learned by
making inferences from already existing factual knowledge. ACT* supports three fundamental types of
learning: generalization, in which productions become broader in their range of application, discrimination,
in which productions become narrow in their range of application, and strengthening, in which some
productions are applied more often. New productions are formed by the conjunction or disjunction of
existing productions.
Conceptual Model
Source: Anderson (1976).
Favorite Methods
Experimental research and Computational simulations.
Scope and Application
ACT* can explain a wide variety of memory effects as well as account for higher order skills such as geometry proofs,
programming and language learning (see Anderson, 1983; 1990). ACT* has been the basis for intelligent tutors (Anderson,
Boyle, Farrell &Reiser, 1987).
Example
One of the strengths of ACT is that it includes both proposition and procedural representation of knowledge as well as
accounting for the use of goals and plans. For example, here is a production rule that could be used to convert declarative
sentences into a question:
IF the goal is to question whether the proposition (LVrelationLVagentLVobject) is true THEN set as subgoals
1. to plan the communication (LVrelationLVagentLVobject)
2. to move the first word in the description of LVrelation to the beginning of the sentence
3. to execute the plan
This production rule could be used to convert the sentence: "The lawyer is buying the car." into the question: "Is the lawyer
buying the car?"
From Ebbinghaus onward psychology has seen an enormous amount of research invested in the study of learning and
memory. This research has produced a steady stream of results and, with a few "mini-revolutions" along the way, a steady
increase in our understanding of how knowledge is acquired, retained, retrieved, and utilized. Throughout this history there
has been a concern with the relationship of this research to its obvious application to education. The first author has written
two textbooks (Anderson, 1995a, 1995b) summarizing some of this research. In both textbooks he has made efforts to
identify the implications of this research for education. However, he left both textbooks feeling very dissatisfied -- that the
intricacy of research and theory on the psychological side was not showing through in the intricacy of educational
application. One finds in psychology many claims of relevance of cognitive psychology research for education. However,
these claims are loose and vague and contrast sharply with the crisp theory and results that exist in the field.
To be able to rigorously understand what the implications are of cognitive psychology research one needs a rigorous theory
that bridges the gap between the detail of the laboratory experiment and the scale of the educational enterprise. This chapter
is based on the ACT-R theory (Anderson, 1993, 1996) which has been able to explain learning in basic psychology
experiments and in a number of educational domains. ACT-R has been advertised as a "simple theory of learning and
cognition". It proposes that complex cognition is composed of relatively simple knowledge units which are acquired
according to relatively simple principles. Human cognition is complex but this complexity reflects complex composition of the
basic elements and principles just as a computer can produce complex aggregate behavior from simple computing elements.
The ACT-R perspective places a premium on the practice which is required to learn permanently the components of the
desired competence. The ACT-R theory claims that to learn a complex competence each component of that competence
must be mastered. It is a sharp contrast to many educational claims, supposedly based in cognitive research, that there are
moments of insight or transformations when whole knowledge structures become reorganized or learned. In contrast, ACT-R
implies that there is no “free lunch” and each piece of knowledge requires its own due of learning. Given the prevalence of
the “free lunch myth” we will endeavor to show that it is not true empirically and to explain why it can not be true within the
ACT-R theory.
This chapter will have the following organization. First we will describe the ACT-R theory and its learning principles. In the
light of this theory, we will identify what we think are the important implications of psychological research for education. We
will also address the issue of why so much of the research on learning and memory falls short of significant educational
application. We will devote special attention to the issues of insight, learning with understanding, and transfer which are part
of the free lunch myth. Finally, we will describe how we have tried to bring the lessons of this analysis to bear in the design of
our cognitive tutors (Anderson, Boyle, Corbett, & Lewis, 1990; Anderson, Corbett, Koedinger, & Pelletier, 1995).
The ACT-R Theory
The ACT-R theory admits of three basic binary distinctions. First, there is a distinction between two types of knowledge -declarative knowledge of facts and procedural knowledge of how to3 do various cognitive tasks. Second, there is the
distinction between the performance assumptions about how ACT-R deploys what it knows to solve a task and the learning
assumptions about how it acquires new knowledge. Third, there is a distinction between the symbolic level in ACT-R which
involves discrete knowledge structures and a sub-symbolic level which involves neural-like activation-based processes that
determine the availability of these symbolic structures. We will first describe ACT-R at the symbolic level. A symbolic-level
analysis of the knowledge structures in a domain corresponds basically to a task analysis of what needs to be learned in that
domain. However, as we will see, the availability of these symbolic structures depends critically on the subsymbolic
processes.
Declarative and Procedural Knowledge
Declarative knowledge reflects the factual information that a person knows and can report. According to ACT-R declarative
knowledge is represented as a network of small units of primitive knowledge called chunks. Figure 1 is a graphical display of
a chunk encoding the addition fact that 3+4=7 and some of its surrounding facts. These are some of the many facts that a
child might have involving these numbers. Frequently, one encounters the question “What does it mean to understand 3 or to
understand numbers in general?” The answer in ACT-R is quite definite on this matter: Understanding involves a large
number of declarative chunks like those in Figure 1 plus a large number of procedural units which determine how this
knowledge is used. According to the ACT-R theory, understanding requires nothing more or less than such a set of
knowledge units. Understanding of a concept results when we have enough knowledge about the concept that we can
flexibly solve significant problems involving the concept.
Procedural knowledge, such as mathematical problem-solving skill, is represented by a large number of rule-like units called
productions. Production rules are condition-action units which respond to various problem-solving conditions with specific
cognitive actions. The steps of thought in a production system correspond to a sequence of such condition-action rules
which execute or (in the terminology of production systems) fire. Production rules in ACT-R specify in their condition the
existence of specific goals and often involve the creation of subgoals. For instance, suppose a child was at the point
illustrated below in the solution of a multi-column addition problem:
534
+248
2
Focused on the tens column the following production rule might apply taken from the ACT-Rsimulation of multi-column
addition in Anderson (1993): IF the goal is to add n1 and n2 in a column and n1 + n2 = n3 THEN set as a subgoal to write n3
in that column
This production rule specifies in its condition the goal of working on the tens column and involves a retrieval of a declarative
chunk like the 3+4=7 fact in Figure 1. In its action it creates a subgoal which might involve things like processing a carry. It is
many procedural rules like this along with the chunks which in total produce what we recognize as competence in a domain
like mathematics.
ACT-R
ACT-R (pronounced act-ARE: Adaptive Control of Thought—Rational) is a cognitive architecture mainly developed by John
Robert Anderson at Carnegie Mellon University. Like any cognitive architecture, ACT-R aims to define the basic and
irreducible cognitive and perceptual operations that enable the human mind. In theory, each task that humans can perform
should consist of a series of these discrete operations.
Most of the ACT-R basic assumptions are also inspired by the progress of cognitive neuroscience, and ACT-R can be seen
and described as a way of specifying how the brain itself is organized in a way that enables individual processing modules to
produce cognition.
Inspiration[edit]
ACT-R has been inspired by the work of Allen Newell, and especially by his lifelong championing the idea of unified theories
[1]
as the only way to truly uncover the underpinnings of cognition. In fact, John Anderson usually credits Allen Newell as the
major source of influence over his own theory.
What ACT-R looks like[edit]
Like other influential cognitive architectures (including Soar, CLARION, and EPIC), the ACT-R theory has a computational
implementation as an interpreter of a special coding language. The interpreter itself is written in Lisp, and might be loaded
into any of the most common distributions of the Lisp language.
This means that any researcher may download the ACT-R code from the ACT-R website, load it into a Lisp distribution, and
gain full access to the theory in the form of the ACT-R interpreter.
Also, this enables researchers to specify models of human cognition in the form of a script in the ACT-R language. The
language primitives and data-types are designed to reflect the theoretical assumptions about human cognition. These
assumptions are based on numerous facts derived from experiments in cognitive psychology and brain imaging.
Like a programming language, ACT-R is a framework: for different tasks (e.g., Tower of Hanoi, memory for text or for list of
words, language comprehension, communication, aircraft controlling), researchers create "models" (i.e., programs) in ACTR. These models reflect the modelers' assumptions about the task within the ACT-R view of cognition. The model might then
be run.
Running a model automatically produces a step-by-step simulation of human behavior which specifies each individual
cognitive operation (i.e., memory encoding and retrieval, visual and auditory encoding, motor programming and execution,
mental imagery manipulation). Each step is associated with quantitative predictions of latencies and accuracies. The model
can be tested by comparing its results with the data collected in behavioral experiments.
In recent years, ACT-R has also been extended to make quantitative predictions of patterns of activation in the brain, as
detected in experiments with fMRI. In particular, ACT-R has been augmented to predict the shape and time-course of
the BOLD response of several brain areas, including the hand and mouth areas in the motor cortex, the left prefrontal cortex,
the anterior cingulate cortex, and thebasal ganglia.
Brief outline[edit]
ACT-R's most important assumption is that human knowledge can be divided into two irreducible kinds of
representations: declarativeand procedural.
Within the ACT-R code, declarative knowledge is represented in the form of chunks, i.e. vector representations of individual
properties, each of them accessible from a labelled slot.
Chunks are held and made accessible through buffers, which are the front-end of what are modules, i.e. specialized and
largely independent brain structures.
There are two types of modules:
Perceptual-motor modules, which take care of the interface with the real world (i.e., with a simulation of the
real world). The most well-developed perceptual-motor modules in ACT-R are the visual and the manual
modules.
Memory modules. There are two kinds of memory modules in ACT-R:
Declarative memory, consisting of facts such as Washington, D.C. is the capital of United
States, France is a country in Europe, or 2+3=5
Procedural memory, made of productions. Productions represent knowledge about how we do
things: for instance, knowledge about how to type the letter "Q" on a keyboard, about how to
drive, or about how to perform addition.
All the modules can only be accessed through their buffers. The contents of the buffers at a given moment in time represents
the state of ACT-R at that moment. The only exception to this rule is the procedural module, which stores and applies
procedural knowledge. It does not have an accessible buffer and is actually used to access other module's contents.
Procedural knowledge is represented in form of productions. The term "production" reflects the actual implementation of
ACT-R as aproduction system, but, in fact, a production is mainly a formal notation to specify the information flow from
cortical areas (i.e. the buffers) to the basal ganglia, and back to the cortex.
At each moment, an internal pattern matcher searches for a production that matches the current state of the buffers. Only
one such production can be executed at a given moment. That production, when executed, can modify the buffers and thus
change the state of the system. Thus, in ACT-R, cognition unfolds as a succession of production firings.
The symbolic vs. connectionist debate[edit]
In the cognitive sciences, different theories are usually ascribed to either the "symbolic" or the "connectionist" approach to
[2]
cognition. ACT-R clearly belongs to the "symbolic" field and is classified as such in standard textbooks and collections. Its
entities (chunks and productions) are discrete and its operations are syntactical, that is, not referring to the semantic content
of the representations but only to their properties that deem them appropriate to participate in the computation(s). This is
seen clearly in the chunk slots and in the properties of buffer matching in productions, both of which function as standard
symbolic variables.
Members of the ACT-R community, including its developers, prefer to think of ACT-R as a general framework that specifies
how the brain is organized, and how its organization gives birth to what is perceived (and, in cognitive psychology,
investigated) as mind, going beyond the traditional symbolic/connectionist debate. None of this, naturally, argues against the
classification of ACT-R as symbolic system, because all symbolic approaches to cognition aim to describe the mind, as a
product of brain function, using a certain class of entities and systems to achieve that goal.
A common misunderstanding suggests that ACT-R may not be a symbolic system because it attempts to characterize brain
function. This is incorrect on two counts: First, because all approaches to computational modeling of cognition, symbolic or
otherwise, must in some respect characterize brain function, because the mind is brain function. And second, because all
such approaches, including connectionist approaches, attempt to characterize the mind at a cognitive level of description
[3]
and not at the neural level, because it is only at the cognitive level that important generalizations can be retained.
Further misunderstandings arise because of the associative character of certain ACT-R properties, such as chunks
spreading activation to each other, or chunks and productions carrying quantitative properties relevant to their selection.
None of these properties counter the fundamental nature of these entities as symbolic, regardless of their role in unit
selection and, ultimately, in computation.
Theory vs. implementation, and Vanilla ACT-R[edit]
The importance of distinguishing between the theory itself and its implementation is usually highlighted by ACT-R
developers.
In fact, much of the implementation does not reflect the theory. For instance, the actual implementation makes use of
additional 'modules' that exist only for purely computational reasons, and are not supposed to reflect anything in the brain
(e.g., one computational module contains the pseudo-random number generator used to produce noisy parameters, while
another holds naming routines for generating data structures accessible through variable names).
Also, the actual implementation is designed to enable researchers to modify the theory, e.g. by altering the standard
parameters, or creating new modules, or partially modifying the behavior of the existing ones.
Finally, while Anderson's laboratory at CMU maintains and releases the official ACT-R code, other alternative
[4]
implementations of the theory have been made available. These alternative implementations include jACT-R (written
in Java by Anthony M. Harrison at theNaval Research Laboratory) and Python ACT-R (written in Python by Terrence C.
[5]
Stewart and Robert L. West at Carleton University, Canada).
[6]
Similarly, ACT-RN (now discontinued) was a full-fledged neural implementation of the 1993 version of the theory. All of
these versions were fully functional, and models have been written and run with all of them.
Because of these implementational degrees of freedom, the ACT-R community usually refers to the "official", lisp-based,
version of the theory, when adopted in its original form and left unmodified, as "Vanilla ACT-R".
Applications[edit]
Over the years, ACT-R models have been used in more than 700 different scientific publications, and have been cited in
many more.
Memory, attention, and executive control[edit]
The ACT-R declarative memory system has been used to model human memory since its inception. In the course of years, it
has been adopted to successfully model a large number of known effects. They include the fan effect of interference for
[7]
[8]
[9]
associated information, primacy and recency effects for list memory, and serial recall.
ACT-R has been used to model attentive and control processes in a number of cognitive paradigms. These include
[10][11]
[12][13]
[14]
[15]
the Stroop task,
task switching,
the psychological refractory period, and multi-tasking.
Natural language[edit]
A number of researchers have been using ACT-R to model several aspects of natural language understanding and
[16]
[17]
[18]
production. They include models of syntactic parsing, language understanding, language acquisition
and metaphor
[19]
comprehension.
Complex tasks[edit]
[20]
ACT-R has been used to capture how humans solve complex problems like the Tower of Hanoi, or how people solve
[21]
[22]
algebraic equations. It has also been used to model human behavior in driving and flying.
With the integration of perceptual-motor capabilities, ACT-R has become increasingly popular as a modeling tool in human
factors and human-computer interaction. In this domain, it has been adopted to model driving behavior under different
[23][24]
[25][26]
[27]
conditions,
menu selection and visual search on computer application,
and web navigation.
Cognitive neuroscience[edit]
[28]
More recently, ACT-R has been used to predict patterns of brain activation during imaging experiments. In this field, ACT[29]
R models have been successfully used to predict prefrontal and parietal activity in memory retrieval, anterior cingulate
[30]
[31]
activity for control operations, and practice-related changes in brain activity.
Education[edit]
[32][33]
ACT-R has been often adopted as the foundation for cognitive tutors.
These systems use an internal ACT-R model to
mimic the behavior of a student and personalize his/her instructions and curriculum, trying to "guess" the difficulties that
students may have and provide focused help.
Such "Cognitive Tutors" are being used as a platform for research on learning and cognitive modeling as part of the
Pittsburgh Science of Learning Center. Some of the most successful applications, like the Cognitive Tutor for Mathematics,
are used in thousands of schools across the United States.
Brief history[edit]
Early years: 1973-1990[edit]
ACT-R is the ultimate successor of a series of increasingly precise models of human cognition developed by John R.
Anderson.
Its roots can be backtraced to the original HAM (Human Associative Memory) model of memory, described by John R.
[34]
[35]
Anderson andGordon Bower in 1973. The HAM model was later expanded into the first version of the ACT theory. This
was the first time the procedural memory was added to the original declarative memory system, introducing a computational
[36]
dichotomy that was later proved to hold in human brain. The theory was then further extended into the ACT* model of
[37]
human cognition.
Integration with rational analysis: 1990-1998[edit]
In the late eighties, Anderson devoted himself to exploring and outlining a mathematical approach to cognition that he
[38]
named Rational Analysis. The basic assumption of Rational Analysis is that cognition is optimally adaptive, and precise
[39]
estimates of cognitive functions mirror statistical properties of the environment. Later on, he came back to the
development of the ACT theory, using the Rational Analysis as a unifying framework for the underlying calculations. To
highlight the importance of the new approach in the shaping of the architecture, its name was modified to ACT-R, with the
[40]
"R" standing for "Rational"
In 1993, Anderson met with Christian Lebiere, a researcher in connectionist models mostly famous for developing with Scott
[41]
Fahlmanthe Cascade Correlation learning algorithm. Their joint work culminated in the release of ACT-R 4.0. Thanks to
Mike Byrne (now atRice University), version 4.0 also included optional perceptual and motor capabilities, mostly inspired
from the EPIC architecture, which greatly expanded the possible applications of the theory.
Current developments 1998-present[edit]
After the release of ACT-R 4.0, John Anderson became more and more interested in the underlying neural plausibility of his
life-time theory, and began to use brain imaging techniques pursuing his own goal of understanding the computational
underpinnings of human mind.
The necessity of accounting for brain localization pushed for a major revision of the theory. ACT-R 5.0 introduced the
concept of modules, specialized sets of procedural and declarative representations that could be mapped to known brain
[42]
systems. In addition, the interaction between procedural and declarative knowledge was mediated by newly introduced
buffers, specialized structures for holding temporarily active information (see the section above). Buffers were thought to
reflect cortical activity, and a subsequent series of studies later confirmed that activations in cortical regions could be
successfully related to computational operations over buffers.
A new version of the code, completely rewritten, was presented in 2005 as ACT-R 6.0. It also included significant
improvements in the ACT-R coding language.
Spin Offs[edit]
The long development of the ACT-R theory gave birth to a certain number of parallel and related projects.
The most important ones are the PUPS production system, an initial implementation of Anderson's theory, later abandoned;
[6]
and ACT-RN, a neural network implementation of the theory developed by Christian Lebiere.
Lynne Reder, also at Carnegie Mellon University, developed in the early nineties SAC, a model of conceptual and perceptual
aspects of memory that shares many features with the ACT-R core declarative system, although differing in some
assumptions.
1 Definition
John R. Anderson's etal.s Adaptive Control of Thought (ACT*) theories are human information processing and knowledge
representation theories.
ACT theory started out in the Simon-Newell tradition, i.e. as a purely symbolic model of human thought and memory. The
latest version is Adaptive control of thought-rational (ACT-R Version 6) (Anderson et al., 2004) and incorporates more recent
ideas about embodyment (perception and action) and subsymbolic processes.
2 Overview
Related to the distinction of declarative vs. procedural knowledge, the critical atomic components of cognition and human
memory are identified as chunks and productions. According to Yates (2007:32), Anderson (1996) claims the following: “
All that there is to intelligence is the simple accrual and tuning of many small units of knowledge that in total produce
complex cognition. The whole is no more than the sum of its parts, but it has a lot of parts. (p. 356).”
According to Yates (2007:33):
Procedural knowledge consists of condition-action (IF-THEN) pairs called productions which are activated
according to rules relating to a goal structure (Anderson, 1983). Within the ACT framework, all knowledge is initially
declarative and is interpreted by general procedures. Productions, then, connect declarative knowledge with behavior.
Procedural knowledge represents "how to do things." It is knowledge that is displayed in our behavior, but that we do not
hold consciously (Anderson &Lebiere, 1998). As a task is performed, interpretive applications are gradually replaced with
productions that perform the task directly, a process called proceduralization. For example, rehearsing how to manually shift
gears in a car is gradually replaced by a production that recognizes and executes the production. In other words, explicit
declarative knowledge is replaced by direct application of procedural knowledge (Anderson, 2005). Sequences of
productions may be combined into a single production, a process called composition. Together, proceduralization and
composition are called knowledge compilation, which creates task-specific productions during practice. The process of
proceduralization affects working memory by reducing the load resulting from information being retrieved from long-term
memory.
See production system and Soar for some technical background.
According to ACT*, all knowledge begins as declarative information; procedural knowledge is learned by making inferences
from already existing factual knowledge. ACT* supports three fundamental types of learning: generalization, in which
productions become broader in their range of application, discrimination, in which productions become narrow in their range
of application, and strengthening, in which some productions are applied more often. New productions are formed by the
conjunction or disjunction of existing productions. (Kearsley: 1994)
Summary of ACT-R (Anderson et al. 2004).
1.
There are multiple independent modules whose information processing is encapsulated.
2.
The modules can place chunks reflecting their processing in their buffers and the production
system can detect when critical patterns are satisfied among these chunks.
3.
From those productions whose conditions are satisfied a single production will be selected at any
time and fire, leading to updates to various buffers that in turn can trigger information
processing in their respective modules.
4.
While chunks and productions are the symbolic components of the system reflecting its overall
information flow, chunks have subsymbolic activations and productions have subsymbolic
utilities that control which chunks and productions get used.
5.
Learning can involve either acquiring new chunks and productions or tuning their subsymbolic
parameters.
6.
These processes are stochastic and take place in real time.
3 ACT as modeling framework
“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” (About, retrieved 11:05, 16 November 2007 (MET)).
4 ACT theory in education
ACT* theory can explain a range of learning types and therefore influence instructional design models. It also is popular in
research onintelligent tutoring systems since the ACT* is a model of a cognitive architecture embedded in a
modeling/programming language. As such it can be used to model learners, e.g. "understand" what difficulties they might
have.
(This section needs to be expanded a lot ...)
Overview
ACT-R is a model of the human cognitive process developed and used by cognitive psychologists, which can
be applied to HCI. It is an acronym for "The Adaptive Control of Thought - Rational". While it is often referred to as "the ACTR theory", it is not properly considered a theory of cognition, but rather a cognitive architecture that can accommodate
different theories. The scope of ACT-R is greater than the scope of any particular theory, and multiple (possibly competing)
theories can fit within the framework of ACT-R. It was developed to model problem solving, learning and memory. ACT-R is
generally used by researchers in cognitive psychology, but researchers have also found applications in HCI.
Production rules
A fundamental characteristic of ACT-R is that it is a production system theory. The basic premise of a
production system theory is that a cognitive skill is composed of conditional statements known as production rules. A
production rule is a statement that describes an action which should be taken if a condition is met, sometimes referred to as
a condition-action pair. For example:
IF the goal is to classify a shape
and the shape has four equal sides
THEN classify the shape as a square.
Cognitive tasks are achieved by stringing together production rules, and applying them to working memory.
Such a collection of production rules is referred to simply as a production. When a production rule is applied, it is said to fire.
Principles
In ACT-R, there are two different categories of long-term memory: declarative and procedural. Declarative
memory consists of facts such as "Annapolis is the capital of Maryland", "A square has four equal sides", or "8*7=56".
Procedural memory consists of our knowledge of how to do things, though we may not be able to verbalize how we are able
to do these things. Examples of procedural knowledge include our ability to drive a car or speak English. Declarative
knowledge is represented in ACT-R by units called chunks. Procedural knowledge is represented by productions, which are
collections of production rules. ACT-R defines a syntax to represent chunks and productions. An ACT-R model can be
represented as a computer program in the LISP programming language, and can be executed. In this syntax, chunks have a
schema-like representation containing an "isa" field specifying the category of knowledge, and additional fields to encode the
knowledge. Below is an encoding of the fact "8*7=56"
fact8*7
isa
multiplication-fact
multiplicand1
eight
multiplicand2
seven
product
fifty-six
Below is an encoding of the production rules for counting from one number from another. It is taken from
the ACT-R Research Groupwebsite.
(P increment
=goal>
ISA
count-from
number
=num1
=retrieval>
ISA
count-order
first
=num1
second
=num2
==>
=goal>
number
=num2
+retrieval>
ISA
count-order
first
=num2
)
Within this production rules paradigm, cognitive tasks are performed by assembling production rules by
setting goals, and by reading and writing to working memory (sometimes referred to as buffers). Goals (and subgoals) are
represented on a structure called the goal stack.
Two other important concepts in ACT-R are pattern matching and conflict resolution. Pattern matching is the
process which determines if a production's conditions are met by the current state of working memory. Conflict resolution is
the process that determines which production should be applied if several production rules are applicable.
ACT-R models are defined on two levels of abstraction: the symbolic level and the subsymbolic level. The
symbolic level is concerned with productions and chunks as described above. These high-level concepts are implemented
by a subsymbolic structure, which consists of a collection of massively parallel processes which are modeled by a set of
mathematical equations. These subsymbolic elements affect the high-level chunks and productions. They can be used to
determine which production to select for execution, and they determine the speed at which information can be retrieved from
declarative memory. They are also responsible for most of the learning processes in ACT-R. The ideal is that this
subsymbolic system accurately models the neurological information processing units of the human brain.
Scope and Application
Since ACT-R is a cognitive architecture, it covers a wide range of human cognitive tasks, focusing on learning
and problem solving. It has been previously applied to modeling such tasks as solving the Tower of Hanoi puzzle, memory
for text or for lists of words, language comprehension, communication and aircraft controlling. To develop an ACT-R model,
one must add domain-specific knowledge to the ACT-R architecture.
Examples
ACT-R models tend to be quite large for all but the most non-trivial of tasks. A prototypical example is the ACTR model for solving the standard Tower of Hanoi problem. This example can be found at the ACT-R research group website.
ACT-R has only recently been applied to HCI. Many of these applications are at a preliminary "proof-ofconcept" stage. Byrne (1999) used ACT-R (specifically, ACT-R/PM) to model random menu selection. Users searched for a
target item on a menu, timings were recorded and compared to an ACT-R model.
Another interesting example of the use of ACT-R applied specifically to HCI is given by Ritter et al (2002). They
suggest the use of ACT-R/PM to design a Cognitive Model Interface Evaluation (CMIE) tool. Such a tool can display a user
interface, run a cognitive model to interact with the interface, provide display facilities for model traces, and predict
performance. They are currently developing a prototype system.