Transcript The Evolution of General Intelligence: The Roles of

The Evolution of General Intelligence:
The Roles of Working Memory and Analogical Reasoning
in Solving Novel Problems



Kevin MacDonald
Department of Psychology
California State University–Long Beach

Evolution, Domain Specificity, and DomainGenerality
• An important aspect of human evolution is the need to
adapt to recurrent environmental features (physical space,
numerosity) and recurrent social problems (e.g., mating,
attaining social status).
• When the environment presents recurrent problems, the
optimal solution is to develop domain-specific cognitive
and psychological mechanisms specialized to handle
specific types of input and generate certain types of
solutions.
• Evolutionary Psychology sees the mind as exclusively or
at least predominantly composed of these mechanisms.
Evolution, Domain Specificity, and DomainGenerality
• Domain-specific mechanisms solve the frame problem—
the problem of assembling task relevant and contentrelevant solutions
• Humans could not have evolved as nothing more than a
generalized fitness maximizer or a general purpose
problem solver.
• Domain-general mechanisms will always be weaker than
domain-specific mechanisms for dealing with recurrent
adaptive problems.
THE EEA ALSO PRESENTED NON-RECURRENT
PROBLEMS BEST SOLVED WITH DOMAINGENERAL MECHANISMS
• Rapid radiation of humans resulted in recurrent
situations of novelty and complexity
(unpredictability) due to rapid ecological changes.
• Evolution of “Adaptive Flexibility” and
encephalization associated with environmental
oscillations. (R. Potts (Variability selection in
Hominid evolution. Evolutionary Anthropology, 7 81–
796, 1998)
• Animals must deal with novelty. Paradigm: Must find
food when usual path to food is blocked or combine
information from several systems in order to solve
problem. Animal g factor
How Domain General Mechanisms Can Evolve




The Satisfaction of Evolved Motivational Dispositions (hunger,
sex, love, safety, social status) need not be achieved via
adaptations sensitive to environmental conditions that were
recurrent in the EEA.
Motivational mechanisms may be thought of as a set of
psychological desires. How we achieve these desires is massively
underspecified.
Motivational systems like hunger (including the ability to know
when hunger is assuaged) enable the evolution of any cognitive
mechanism, no matter how opportunistic, flexible, or domaingeneral, that is able to solve the problem.
Hunger could be alleviated by discovering a novel contingency
(operant conditioning), by observing others (social learning), or by
developing a novel plan requiring explicit representations of
events and a great deal of working memory—general
intelligence.
Level 1
EVOLVED MOTIVE DISPOSITIONS
(Domain-Specific Mechanisms)
Level 2
PERSONAL STRIVINGS
(Direct Psychological Effects of
Domain-Specific Mechanisms)
Level 3
CONCERNS, PROJECTS, TASKS
(May Utilize Domain-General
Mechanisms)
Level 4
SPECIFIC ACTION UNITS
(May Utilize Domain-General
Mechanisms)
____________________________________________________
EXAMPLE:
Evolved Motive Disposition: INTIMACY (or Social status, safety, sexual
gratification)
Personal Striving: Intimate relationship with a particular person
Concern, Project, Task: Arrange Meeting, Improve appearance, Get
promotion
Action Units: Find phone number, Begin dieting, Work on weekends
Figure 2. Hierarchical model of motivation showing relationships between domainspecific and domain-general mechanisms (after Emmons, 1989).
From: MACDONALD, K. B. (1991). A PERSPECTIVE ON DARWINIAN PSYCHOLOGY:
THE IMPORTANCE OF DOMAIN-GENERAL MECHANISMS, PLASTICITY, AND INDIVIDUAL
DIFFERENCES. ETHOLOGY AND SOCIOBIOLOGY, 12, 449–480.
Problem Solving is an Opportunistic, GoalDirected Activity
• Being Restricted to Adaptations Linking
Environmental Events Recurrent in the EEA with
Achieving EMD’s is a Non-Necessity. We can come
up with new ways to solve old problems.
• Humans are Flexible Strategizers. “Children reason about
wide-ranging situations and problems for which they have
no special-purpose tools. They bring to bear varied
processes and strategies, gradually coming through
experience to select those that are most effective. . . .
Young bricoleurs . . . make do with whatever cognitive
tools are at hand” (J. S. Deloache, K. F. Miller & S. L.
Pierroutsakos, (1998)
Hypothesis: Function of general intelligence is the
attainment of evolutionary goals in unfamiliar,
novel, or unpredictable conditions characterized by
a minimal amount of prior knowledge
• We propose that there are a variety of functionally
domain-general problem solving mechanisms
designed to respond adaptively (I.e., facilitate
attaining evolutionary goals) to problems that were
not sufficiently recurrent to result in the evolution of
dedicated, domain-specific systems. g is a
functional competency that includes working
memory, inhibition, and abstraction (decontextualization).
Intelligence and Novelty

Fluid intelligence: “Gf reasoning abilities consist
of strategies, heuristics, and automatized
systems that must be used in dealing with novel
problems, educing relations, and solving
inductive, deductive, and conjunctive reasoning
tasks” Horn, J. L., & Hofer, S. M. (1992).

Carl Bereiter: Intelligence is “what you use when
you don’t know what to do.”
Intelligence involves conscious problem
solving and is relatively slow compared to
the unconscious, automatic processing
characteristic of modular, domain-specific
systems.
Working memory is critical: “g-loaded tasks
require high working memory = becoming aware
of information, discriminating between different
bits of information, retaining such awarenesses
and discriminations over short periods of time in
performing various kinds of tasks” (Horn & Hofer,
1992, p. 62).
Working memory, analogical reasoning
and IQ are intercorrelated.
Analogical Reasoning
• Correlations range from .68 to .84 between tests of
general intelligence and tests of analogical
reasoning (Spearman, 1927, The Abilities of Man;
Sternberg 1977; see also Sternberg & Gardner, 1982).
• Analogies, such as “sound is like a water wave,”
involve transferring information across conceptual
domains (Chiappe, 1998, 2000; Gentner & Holyoak,
1997; Holyoak & Thagard, 1989, 1995, 1997).
• Source: Water Wave
• Target: Sound
Analogical Reasoning is Domain General



Analogies establish relevant similarities between a source
domain (e.g., water waves) and a target domain (e.g.,
sound). This allows us to use a familiar situation as a
model for making inferences about an unfamiliar situation
(solving novel problems).
There are no limits on the domains that can be connected
via an analogy.
Science (Solving Novel Problems):
– Huygens: Light and sound
– Darwin: Natural selection and artificial selection
– Kekulé: Benzene molecule and a snake eating it’s tail
– Psychology: The mind and wax tablets, blank slates, steam
engines, telephone networks, and digital computers
Analogical Reasoning is Domain General


Technology:
– Alexander Graham Bell: Ear as model for telephone
– Georges de Mestral: Burrs sticking to dog as model for
velcro:
The Law:
– Precedent-based Reasoning

Political Rhetoric:
– Domino theory of communism
– Hitler = Saddam Hussein

Everyday Conversation:
– “We’re at a crossroads”; “We’re spinning our wheels”
Analogical Reasoning is Unencapsulated

Domain-specific modules are encapsulated. They
respond to a narrow range of information (e.g.,
the face recognition module), but analogical
reasoning utilizes information from widely
disparate areas:

Jerry Fodor (1983, 107): “By definition,
encapsulated systems do not reason
analogically.”
Analogical Reasoning is Unencapsulated

We can compare and contrast virtually any two
concepts that we explicitly represent. Lawyers
can be compared to sharks, junk yard dogs,
snakes, weasels, jackals, carnival barkers,
charlatans, quacks, teddie bears, etc.

Education is a stairway, an obstacle course, a
smorgasbord, a trial by fire, a party, etc.

Providing subjects with analogies from very
different domains facilitates problem solving
(Gick & Holyoak 1980).
Dedre Gentner’s “Structure-Mapping” Theory

Analogies require ability to consciously manipulate explicit
mental representations = meta-representational abilities.

Key similarities are not between attributes of objects but
between relations or relations between relations:

“The key similarities lie in the relations that hold within
domains (e.g., the flow of electrons in an electrical circuit is
analogically similar to the flow of people in a crowded
subway tunnel) rather than in features of individual objects
(e.g., electrons do not resemble people)” (Gentner &
Holyoak 1997, p. 33).
Dedre Gentner’s “Structure-Mapping” Theory

People prefer interpretations that involve establishing
similarities at abstract levels.

Individual relations across domains are brought into
correspondence on the basis of their common role in the
overall causal structure, and we ignore relations that can’t
be put into such causal structures.
Heat Flow and Water Flow Analogy
Heat Flow and Water Flow Analogy

Key relationships:
• FLOW (water, pipe, beaker, vial) corresponds to
FLOW (heat, bar, coffee, ice)
• GREATER [PRESSURE (beaker), PRESSURE (vial)]
corresponds to GREATER [TEMPERATURE (coffee
cup), TEMPERATURE (ice cube)]

Ignore GREATER [DIAM (beaker] and GREATER [DIAM
(vial)] because
it can’t be placed into a
causal structure
Working Memory: Working Memory Linked
both to IQ and to Analogical Reasoning

Analogical reasoning requires a great deal of
conscious mental effort, making substantial use
of the resources of working memory.

Requires both a storage component and an
attention-demanding, processing component —
two hallmarks of working memory tasks.
Working Memory: Working Memory Linked
both to IQ and to Analogical Reasoning



Must activate important elements and relations of
the domains involved while searching for abstract
commonalities between the two.
Must inhibit potentially distracting components of
the domains, such as some of their superficial
features that may not contribute to the final
interpretation of the analogy.
Must keep active the current processing goals
motivating the analogy, and that drive the
mapping process.
Correlations between Verbal Analogies and
Measures of Working Memory

Correlations between Verbal
Analogies and Working memory
capacity tests
•
•
•
•
ABC Numerical Assignment: .54
Digit Span: .36
Mental arithmetic: 43
Alphabet Re-coding: .44
• “Reasoning Ability Is (Little More Than Working-Memory
Capacity ?!”, Kyllonen & Christal (Intelligence 14, 389433, 1990)
Cosmides & Tooby (2002): Intelligence as Local
Contingency/Hyper-Contextualization

Humans solve recurrent problems via domain-specific
modules. Novel problems without cues that have been
recurrent over evolutionary time are solved by a “scope
syntax” that marks certain bits of information as only
locally true or false and includes “a set of procedures,
operators, relationships, and data-handling formats that
regulate the migration of information among subcomponents of the human cognitive architecture” (L.
Cosmides & J. Tooby, Unraveling the enigma of human intelligence:
Evolutionary psychology and the multimodular mind. In R. J. Sternberg & J. C.
Kaufman (Eds.), The Evolution of Intelligence, pp. 145–198. Mahwah, NJ: Lawrence
Erlbaum.)

Implies that intelligence involves “hyper-contextualization”
because it highlights local contingency.
This conflicts with a long history of data showing intelligence
is linked with de-contextualization—Overriding Local
Contingency in favor of abstraction of commonalities
The Terms for the Two Systems Used by a Variety of
Theorists and the Properties of Dual-Process Theories of
Reasoning (Stanovich and West, 2000)
Lower g
Higher g
associative
rule-based
Properties:
holistic
analytic
automatic
controlled
relatively
demanding of cognitive
undemanding of
capacity
cognitive
capacity
relatively fast
relatively slow
acquisition by
acquisition by cultural and
biology,
formal tuition
exposure, and
personal
experience
Task
highly
decontextualized
Construal:
contextualized
personalized
depersonalized
conversational
asocial
and social
Type of
interactional
analytic
Intelligence (conversational
(psychometric IQ)
Indexed:
implicature)
Analogical Reasoning Involves
De-Contextualization




Through representational redescription, patterns and
relationships embedded in a particular domain become
represented more explicitly and more abstractly.
“Information already present in the organism’s
independently functioning, special-purpose
representations, is made progressively available…to other
parts of the cognitive system” (Karmiloff-Smith 1992, pp.
17-18).
The process of abstracting a schema is essentially decontextualization — one “deletes differences between the
analogs while preserving their commonalities” (Holyoak,
1984, p. 208).
Through this process one creates new systems of higherorder relations that can be applied across a wide range of
domains.
Correspondences among Two Convergence Problems and
Their Schema (from Gick & Holyoak, 1983)

Military Problem
– Initial State
• Goal: Use of Army to capture fortress
• Resources: Sufficiently large Army
• Constraint: Unable to send entire army along one road
– Solution Plan: Send small groups along multiple roads
– Outcome: Fortress captured by army
 Radiation Problem
– Initial State
• Goal: Use x-rays to destroy tumor
• Resources: Sufficiently powerful rays
• Constraint: Unable to administer high-intensity rays from
one direction
– Solution Plan: Administer low-intensity rays from multiple
directions
– Outcome: Tumor destroyed by ray
Correspondences among Two Convergence Problems and
Their Schema (from Gick & Holyoak, 1983)

Convergence Schema [Abstract, Decontextualized]
– Initial State
• Goal: Use force to overcome a central target
• Resources: Sufficiently great force
• Constraint: Unable to apply full force along one path
– Solution Plan: Apply weak forces along multiple paths
simultaneously
– Outcome: Central target overcome by force
Conclusion: General Intelligence is a DomainGeneral Adaptation whose Adaptive Function
is to Enable Humans to Solve Novel Problems
and Thereby Attain Ancient Evolutionary
Goals of Survival and Reproduction