|
Fit2 for Connectionist
and Algebraic Modeling
This program allows you to simulate prominent connectionist
and algebraic models of learning and induction.
Social
Connectionism: A Reader and Handbook for Simulations
Author - Frank Van Overwalle
List Price: £59.95 (Hardcover, 456 pages)
ISBN: 9781841696652
ISBN-10: 184169665X
Publisher: Psychology Press
Publication Date: 27/04/2007
Available
for Order
Try out a selected chapter
8 (available as print proof).
Download for free the accompanying
FIT Program or the FIT Exercises described in the book
About the Book
Many of our thoughts and decisions occur without us being conscious
of them taking place; connectionism attempts to reveal the internal
hidden dynamics that drive the thoughts and actions of both individuals
and groups. Connectionist modeling is a radically innovative approach
to theorizing in psychology, and more recently in the field of social
psychology. The connectionist perspective interprets human cognition
as a dynamic and adaptive system that learns from its own direct
experiences or through indirect communication from others.
Social Connectionism offers an overview of the most recent theoretical
developments of connectionist models in social psychology. The volume
is divided into four sections, beginning with an introduction and
overview of social connectionism. This is followed by chapters on
causal attribution, person and group impression formation, and attitudes.
Each chapter is followed by simulation exercises that can be carried
out using the FIT simulation program; these guided exercises allow
the reader to reproduce published results.
Social Connectionism will be invaluable to graduate students and
researchers primarily in the field of social psychology, but also
in cognitive psychology and connectionist modeling.
Contents
- Part I: Basics.
Van Overwalle, Introduction and Overview. Vanhoomissen &
Van Overwalle, Connectionist Basics.
Van Overwalle & Vanhoomissen, Recurrent and Feedforward
Connectionist Networks, and their Emergent Properties.
- Part II: Causal Attribution.
Van Overwalle & Van Rooy, When More Observations are Better
Than Less: A Connectionist Account of the Acquisition of Causal
Strength.
Read & Montoya, An Autoassociative Model of Causal Reasoning
and Causal Learning: Reply to Van Overwalle's (1998) Critique
of Read and Marcus-Newhall (1993).
Van Overwalle, When One Explanation is Enough: A Connectionist
View on the Fundamental Attribution Bias.
- Part III: Person and Group Impression Formation.
Smith & DeCoster, Knowledge Acquisition, Accessibility,
and Use in Person Perception and Stereotyping: Simulation With
a Recurrent Connectionist Network.
Van Overwalle & Labiouse, A Recurrent Connectionist Model
of Person Impression Formation.
Van Rooy, Van Overwalle, Vanhoomissen, Labiouse & French,
A Recurrent Connectionist Model of Group Biases.
Queller & Smith, Subtyping Versus Bookkeeping in Stereotype
Learning and Change: Connectionist Simulations and Empirical
Findings.
- Part IV: Attitudes.
Van Overwalle & Siebler, A Connectionist Model of Attitude
Formation and Change.
- Appendix: FIT Manual
Unique features
of the FIT program
About FIT:
"My students are absolute beginners with respect
to running a simulation, but have mastered the FIT2 program pretty
quickly. Okay, "mastered" may be a slight exaggeration,
but they learned pretty quickly how to use it. The program is a
very valuable tool!" (Frank Siebler)
-
You can directly compare the simulation output
with real observed data from actual experiments (hence its name
FIT). While it is of great importance to test whether a simulated
theoretical model can reproduce actual data, this often is a
tedious job in other programs. In this program, this is the
basic of the program input (although it is possible also to
specify no actual data).
-
In addition, you can automatically search for
the parameter values of the simulated model that best fit with
your actual data.
-
The program allows you to specify different categories
(blocks) of trial and test data, which can be processed in a
random order per category if you wish. Most importantly, each
of these categories can be processed in an order that you specify
in a session. This is allows you to follow actual experimental
procedures or imaged learning histories in detail, without complicated
script writing. It is even possible to specify several session
categories, so that you can test different learning histories
for the same data.
-
You specify the data input (trial and session
categories) in a user’s friendly data grid, which is very
similar to common spreadsheets like Excel. The simulated output
is also given as grid data, and can be visually inspected by
graphs, or can be exported to other programs.
The following models
are currently available
-
Feedforward Connectionist Models: Feedforward
(McClelland and Rumelhart, 1988), Configural (Pearce, 1994),
BackPropagation (McClelland & Rumelhart, 1988), Simple Recurrent
Net (Elman, 1990)
-
Recurrent Connectionist Models: Linear Auto-associator
(McClelland and Rumelhart, 1988), BSB Linear Auto-associator
(McClelland and Rumelhart, 1988), Non-Linear Auto-associator
(McClelland and Rumelhart, 1988), DiscoNet with hidden layer
(Labiouse, 1999)
-
Algebraic Models: Modular Probabilistic (Cheng
and Novick, 1990), Modular ANOVA (Försterling, 1989), Updating
(Busemeyer, 1991; Hogarth & Einhorn, 1992), Evidential Evaluation
(White, 1998)
-
Associative Models (predecessors of connectionist
models): Modular Associative (Rescorla and Wagner, 1972), Configural-Cue
Associative (Gluck and Bower, 1988), Dimensionalized Configural-Cue
(Gluck and Bower, 1988)
Screen Shots of
the Fit program
This is a view on the spreadsheet-like input from
the program and some output (click to enlarge)
This is a view on the graphical interface to visualize
the simulated data (broken lines) and the fit with the human data
(full lines). (Click to enlarge)
Download of
the Fit Program and accompanying files (free-ware)
-
Download the FIT Program (for Windows only) and upon completion,
the program will be installed. This FIT software is the sole
responsibility of Frank van Overwalle and is not affiliated
with Psychology Press, inc.
-
-
Download PowerPoint
presentations on some of our publications in which the FIT program
was applied.
-
For technical support issues, please contact
Frank van Overwalle directly at Frank.VanOverwalle@vub.ac.be or
by choosing the menu Tools | Email in the FIT
program.
About the author: Van
Overwalle, Frank
Vrije Universiteit Brussel
Department of Psychology
Pleinlaan 2
B - 1050 Brussel
Belgium
phone : + 32-2-629 25 18
fax : + 32-2-629 24 89
email : Frank.VanOverwalle@vub.ac.be
 My
major research interest is currently on connectionist
models of important domains in social cognition: causal
attribution, group biases, person impression formation
and attitude formation and change (including cognitive
dissonance). I conducted simulations on representative
findings from the literature in these domains using common
network architectures and processing parameters in order
to develop a general and unified process model of these
judgments in social cognition. In addition, I also devised
experiments to test some specific predictions that emerged
from these network simulations and that sometimes contradict
currently held beliefs on how these judgments are made.
Selection of
Publications with FIT
-
Van Overwalle, F. & Heylighen, F.
(2006). Talking Nets: A Multi-Agent Connectionist Approach
to Communication and Trust between Individuals. Psychological
Review, 113, 606-627.
-
-
Van Overwalle, F. (2004). Multiple
Person Inferences: A View of a Connectionist Integration.
In H. Bowman & C. Labiouse (Eds.), Connectionist models
of cognition and perception II: Proceedings of the Eighth
Neural Computation and Psychology Workshop. London, UK:
World Scientific
-
-
Van Overwalle, F. (2003) Acquisition
of Dispositional Attributions: Effects of Sample Size
and Covariation. European Journal of Social Psychology,
33, 515-533.
-
Jordens, K. & Van Overwalle,
F. (2003). Connectionist Modelling of Attitudes and
Cognitive Dissonance. In G. Haddock & G. Maio (Eds.)
Contemporary perspectives on the psychology of attitudes.
London: Psychology Press.
-
Van Rooy, D, Van Overwalle, F., Vanhoomissen,
T., Labiouse, C & French, R. (2003). A recurrent
connectionist model of group biases. Psychological Review,
110, 536-563.
-
-
-
Van Overwalle, F. & Timmermans,
B. (2001) Learning about an Absent Cause: Discounting
and Augmentation of Positively and Independently Related
Causes In French, R.M., & Sougné, J.P. (Eds.)
Connectionist Models of Learning, Development and Evolution:
Proceedings of the Sixth Neural Computation and Psychology
Workshop, Liege, Belgium, 16-18 September 2000. Springer
-
Van Overwalle, F. & Van Rooy, D. (2001).
When more observations are better than less : A connectionist
account of the acquisition of causal strength. European Journal of Social Psychology, 31, 155-175.
Abstracts
Talking Nets: A Multi-Agent Connectionist Approach
to Communication and Trust between Individuals.
A multi-agent connectionist model is proposed
that consists of a collection of individual recurrent networks
that communicate with each other, and as such is a network
of networks. The individual recurrent networks simulate
the process of information uptake, integration and memorization
within individual agents, while the communication of beliefs
and opinions between agents is propagated along connections
between the individual networks. A crucial aspect in belief
updating based on information from other agents is the trust
in the information provided. In the model, trust is determined
by the consistency with the receiving agents’ existing
beliefs, and results in changes of the connections between
individual networks, called trust weights. Thus activation
spreading and weight change between individual networks
is analogous to standard connectionist processes, although
trust weights take a specific function. Specifically, they
lead to a selective propagation and thus filtering out of
less reliable information, and they implement Grice’s
(1975) maxims of quality and quantity in communication.
The unique contribution of communicative mechanisms beyond
intra-personal processing of individual networks was explored
in simulations of key phenomena involving persuasive communication
and polarization, lexical acquisition, spreading of stereotypes
and rumors, and a lack of sharing unique information in
group decisions.
A Connectionist Model of Attitude Formation
and Change
This paper discusses a recurrent connectionist
network, simulating empirical phenomena usually explained
by current dual-process approaches of attitudes, thereby focusing
on the processing mechanisms that may underlie both central
and peripheral routes of persuasion. Major findings in attitude
formation and change involving both processing modes are reviewed
and modeled from a connectionist perspective. We use an auto-associative
network architecture with a linear activation update and the
delta learning algorithm for adjusting the connection weights.
The network is applied to well-known experiments involving
deliberative attitude formation as well as the use of heuristics
of length, consensus, expertise and mood. All these empirical
phenomena are successfully reproduced in the simulations.
Moreover, the proposed model is shown to be consistent with
algebraic models of attitude formation (Fishbein & Ajzen,
1975). The discussion centers on how the proposed network
model may be used to unite and formalize current ideas and
hypotheses on the processes underlying attitude acquisition,
and how it can be deployed to develop novel hypotheses in
the attitude domain.
Multiple Person Inferences: A View of a Connectionist
Integration.
This paper provides a connectionist account
of the processes underlying the multiple inference model of
person impression formation proposed by Reeder, Kumar, Hesson-McInnis
and Trafimow (2002). First, in a replication and extension
of one of their main studies, I found evidence for discounting
of trait inferences when facilitating situational forces were
present consistent with earlier causality-based theories,
while at the same time I replicated the lack of discounting
in moral inferences as documented and predicted by Reeder
et al. (2002). Second, to provide an account of how these
different and sometimes contradictory inferences are formed
and integrated in a coherent person impression, I present
a recurrent network model that automatically integrates these
inferences, resulting in a pattern that closely reproduces
the observed data.
Acquisition of Dispositional Attributions: Effects
of Sample Size and Covariation
Two experiments examined whether dispositional
attributions are sensitive to the sample size of the evidence
indicating a given level of covariation between person and
behavior. Participants were given high or low levels of covariation
(i.e., consensus and distinctiveness), and the acquisition
of dispositional attributions was monitored by requesting
dispositional trait ratings at fixed intervals. The results
showed that dispositional attributions were sensitive to sample
size, and increased given more evidence on high person-behavior
covariation while they decreased given more evidence on low
person-behavior covariation. Additional analyses suggested
that in making dispositional inferences (e.g., about the actor),
there was a slight preference for agreement information (e.g.,
low distinctiveness) over difference information (e.g., low
consensus). The effects of sample size are inconsistent with
current statistical or probabilistic models of covariation,
but are in line with connectionist networks using an error-correcting
learning algorithm.
A Recurrent Connectionist Model of Person
Impression Formation
Major findings in impression formation are
reviewed and modeled from a connectionist perspective. The
findings are in the areas of primacy and recency in impression
formation, asymmetric diagnosticity of ability- and morality-related
traits, increased recall for trait-inconsistent information,
assimilation and contrast in priming, and discounting of trait
inferences by situational information. The majority of these
phenomena are illustrated with well-known experiments, and
simulated with an auto-associative network architecture with
linear activation update and using the delta learning algorithm
for adjusting the connection weights. All of the simulations
successfully reproduced the empirical findings. Moreover,
the proposed model is shown to be consistent with earlier
algebraic models of impression formation (Anderson, 1981;
Busemeyer, 1991; Hogarth & Einhorn, 1992). The discussion
centers on how the present model compares to other connectionist
approaches to impression formation and how it may contribute
to a more parsimonious and unified theory of person perception.
Connectionist Modelling of Attitudes
and Cognitive Dissonance
[From the first page...] How are attitudes
represented in human memory, and how are they changed after
direct experiences or messages that contradict earlier opinions?
In this chapter, the question of attitude representation and
change is analysed from a connectionist approach... A key
difference from earlier models is that the connectionist architecture
and processing mechanisms are founded on the neurological
properties of the brain. This allows a view of the mind as
an adaptive learning mechanism that develops an accurate mental
representation of the world. Learning is modelled as a process
of on-line adaptation of existing knowledge to novel information
provided by the environment.
A Recurrent Connectionist Model of Group Biases
Major biases and stereotypes in group judgments
are reviewed and modeled from a recurrent connectionist perspective.
These biases are in the areas of group impression formation
(illusory correlation), group differentiation (accentuation),
stereotype change (dispersed versus concentrated distribution
of inconsistent information), and group homogeneity. All these
phenomena are illustrated with well-known experiments, and
simulated with an auto-associative network architecture with
linear activation update and delta learning algorithm for
adjusting the connection weights. All the biases were successfully
reproduced in the simulations. The discussion centers on how
the particular simulation specifications compare to other
models of group biases and how they may be used to develop
novel hypotheses for testing the connectionist modeling approach
and, more generally, for improving theorizing in the field
of social biases and stereotype change.
An
adaptive Connectionist Model of Cognitive Dissonance.
This article proposes an adaptive connectionist
model that implements an attributional account of cognitive
dissonance. The model represents an attitude as the connection
between the attitude object and behavioral-affective outcomes.
Dissonance arises when circumstantial constraints induce a
mismatch between the model's (mental) prediction and discrepant
behavior or affect. Reduction of dissonance by attitude change
is accomplished through long-lasting changes in the connection
weights using the error-correcting delta learning algorithm.
The model can explain both the typical effects predicted by
dissonance theory as well as some atypical effects (i.e.,
reinforcement effect), using this principle of weight changes
and by giving a prominent role to affective experiences. The
model was implemented in a standard feedforward connectionist
network. Computer simulations showed an adequate fit with
several classical dissonance paradigms (inhibition, initiation,
forced compliance, free choice & misattribution), as well
as novel studies that underscore the role of affect. A comparison
with an earlier constraint satisfaction approach (Shultz &
Lepper, 1996) indicates that the feedforward implementation
provides a similar fit with these human data, while avoiding
a number of shortcomings of this previous model.
How One Cause Discounts or Augments Another: A Connectionist
Account of Causal Competition
We investigated the degree of discounting
and augmentation of a target cause by an alternative cause,
given a varying number of observations on the alternative
cause while holding its degree of covariation constant. Two
experiments showed that more observations of the alternative
cause resulted in greater discounting or augmentation of a
target cause. This sample size effect cannot be explained
by current attribution theories based on statistical notions
or belief updating, but can be accounted for by a connectionist
framework. In addition, we found that the sample size effect
was stronger when the information was presented in a sequential
trial-by-trial format as opposed to a summarized format, but
found no effect of information order. Possible extensions
of statistical models with confidence weights that take account
of sample size were considered and simulated, but none of
them accommodated the data as well as connectionist models.
Learning about an Absent Cause: Discounting
and Augmentation of Positively and Independently Related
Causes
Standard connectionist models of pattern completion
like an auto-associator, typically fill in the activation
of a missing feature with internal input from nodes that are
connected to it. However, associative studies on competition
between alternative causes, demonstrate that people do not
always complete the activation of a missing feature, but rather
actively encode it as missing whenever its presence was highly
expected. Dickinson and Burke's revaluation hypothesis [4]
predicts that there is always forward competition of a novel
cause, but that backward competition of a known cause depends
on a consistent (positive) relation with the alternative cause.
This hypothesis was confirmed in several experiments. These
effects cannot be explained by standard auto-associative networks,
but can be accounted for by a modified auto-associative network
that is able to recognize absent information as missing and
provides it with negative, rather than positive activation
from related nodes.
When
More Observations are Better Than Less: A Connectionist
Account of the Acquisition of Causal Strength
The statistical law of large numbers prescribes
that estimates are more reliable and accurate when based on
a larger sample of observations. This effect of sample size was investigated
on causal attributions.
Subjects received fixed levels of consensus and distinctiveness
covariation, and attributions were measured after a varying
number of trials. Whereas prominent statistical models of
causality (e.g., Cheng & Novick, 1990; Försterling,
1992) predict no effect of sample size, adaptive connectionist
models (McClelland & Rumelhart, 1988) predict that subjects
will incrementally adjust causal ratings in the direction
of the true covariation the more observations are made. In three experiments, sample effects were
found consistent with the connectionist prediction. Possible extensions of statistical models
were considered and simulated, but none of them accommodated
the data as well as connectionist models.
|
|
