|
The Cognitive Lunch talks will be on Wednesdays from 12:10 - 1:25 in the Psychology conference room (PY 128) located behind the main office..
Abstracts
09/08: Kelly Addis
Chaining together serial and free recall
Previous attempts to account for serial and free recall within a single framework are limited in number, and required separate implementations for each task. We present a strength model of
associative chaining (SMAC) designed to fit to key behavioral results from both serial and free recall, in both single--trial and multi--trial paradigms. The model successfully accounts for serial position and output order effects, as well as changes in these effects over learning trials in both recall conditions. Unlike previous theories, SMAC can simultaneously account for both serial and free recall within a single implementation.
09/15: Michael Roberts
Goldstone and Ashpole (2004) examined dynamic group foraging behavior by using
an experimental networked Java platform. I will present two agent-based models
that capture the behavior of the two perceptual conditions in Goldstone and
Ashpole, and I will describe model-inspired explanations for the undermatching
of humans to resources found in the empirical data. Contrary to common theories
in the animal behavior literature, these models also show that competitive
differences among individuals, travel costs, and interference are not necessary
to obtain undermatching of organisms to resources.
09/22: Ami Eidels
Relating Stroop and search phenomena
The Stroop effect is the classic example of human failure to attend
selectively. When presented with color-words written in colors, participants
are slower naming the color of incongruent stimuli(e.g., the word RED written
in green color) than they are naming congruent ones (say, RED in red).
participants seem to be unable to focus selectively on the color.
Common to all classic theories of the Stroop effect is the notion of
interaction between the processing channels of word and color. According to
these theories, some sort of a cross-talk must occur between channels.
Applying Townsend's system factorial technology, I challenge the idea that a
word-color interaction is necessary to produce the behavioral Stroop effect.
09/29: Richard Shiffrin
How many processes are in the brain?
A viewer of psychology from afar would think we were obsessed with binary concepts and dichotomies. Hardly any issue results in more studies, replies, arguments, and controversy than the question: "Are there one or two processes?". In an occasional frenzy of complexity some researchers include the possibility of three processes. This observation applies to almost any topic. In a departure from the usual data driven
talk and model fits, I will discuss this phenomena, asking such questions as: "Does it make sense to ask whether one or two processes are true, given that all our models are at best crude approximations?" "Is there any utility to carrying out our scientific enterprise in such a fashion?" "What role does Occam's Razor and the preference for simpler explanations play in these debates?" "How do we adjudicate these debates?" "How should we adjudicate?" "How do these questions relate to qualitative and quantitative modeling, and to hypothesis testing?" I will illustrate with examples from my own research career, a career filled with such approaches, and replete with examples on every side of the questions.
10/06: Robert Goldstone (presenting research conducted with Andy Jones and Michael Roberts)
Group Path Formation
When people make choices within a group, they are frequently influenced by the choices made by others. We have experimentally explored the general phenomenon of group behavior where an early action facilitates subsequent actions. Our concrete instantiation of this problem is group path formation where people travel between destinations with the travel cost for moving
onto a location inversely related to the frequency with which the location has been visited by others. To empirically study collective path formation, we have developed software that allows upwards of 80 people to simultaneously interact in a virtual environment, viewing moment-by-moment changes in others' movements and the environment. We compare the resulting
paths to formal solutions (Minimal Steiner Trees - MST) and the "Active Walker" model of pedestrian motion from biophysics. There were systematic deviations from beeline pathways in the direction of the MST solution. These deviations showed asymmetries (people took different paths from A to B than they did from B to A), and varied as a function of the topology of the
destinations, the duration of travel, and the absolute scale of the world.
10/20: Shane Mueller and Christoph Weidemann
Alphabetic Letter Similarity
For more than 120 years, psychologists and vision researchers have conducted experiments examining the similarity structure of the complete roman alphabet. We will review some of the conclusions reached by this research, and introduce a new method for measuring the
similarity structure of the alphabet that has not been used previously. The use of this method to measure the visual similarity of letters enables a novel conclusion: that along with factors such as similarity, perceivability, and response biases, identification
accuracy is also affected by discounting of information perceived during the task. We will also discuss our attempts to represent the similarity of the alphabet in a tree form, theoretical motivations for such a representation, and the statistical reliability of some techniques used to generate these trees.
10/27: Jerome Busemeyer
Quantum Dynamics of Human Decision Making
Information processing theories of cognitive psychology (production systems as well as connectionist networks) are based on principles of standard logic and classic probability theory. In this talk, I will introduce ideas from Quantum Information Processing theory, and apply these ideas to divided attention, probability judgments, and decision making under conflict. Quantum reasoning can depart radically from classical reasoning -- for example, according to classic reasoning, the probability of the union of two paths to reach a conclusion must be greater than the probability of each path alone; but Quantum theory allows the probability of each single path to exceed the probability of the union of the two paths. The latter result is not considered 'irrational' from a Quantum point of view. Several empirical examples of these types of counterintuitive violations of classical reasoning will be reviewed and analyzed from a Quantum information processing point of view.
11/03: Adam Sanborn
Seeing Words Without Seeing Features
Visual word recognition is often studied by following the target word with a mask to reduce performance. However, using a mask can change how the word is processed. To explore how well observers see words and their features, the similarity between the two alternatives in a forced choice task was manipulated. Three choice conditions were used: word information only (e.g., PITY – CITY), case information only (e.g., PITY – pity), and both word and case information (e.g., PITY – city). When the target word was not masked, observers were able to effectively use the large visual differences between uppercase and lowercase. But when the target word was masked, observers were more effective using the visually small but meaningful difference between words. Further experiments extend the results to color features, explore nonword and number stimuli, and look at the role of observer strategies.
11/10: Ariane Lambert-Mogiliansky
The logic of Human behavior, Can it be Quantum Logic
Instances of non-commutativity are pervasive in human behavior. In this paper, we suggest that psychological properties such as attitudes, values, preferences may be suitably described in terms of the mathematical formalism of quantum mechanics i.e., quantum logic. We develop a (empirical) notion of measurement and construct a theory relying on a few axioms. An interpretation of these axioms in psychology is proposed and an application to decision theory is suggested.
11/17: Jason Gold
External noise distinguishes exemplar and prototype category representations
How are categories represented in human memory? Some theories, known as ‘exemplar’ models, assume categories are represented by storing all of the individual items we have experienced in memory.
New items are then classified according to their similarities to these stored items. A second class of theories, known as ‘prototype’ models, assume that categories are represented by constructing a series of summary abstractions that are the central tendency of all of the items
we have experienced within each category, and new items are classified according to their similarities to these individual category prototypes. These differences in representation between the models have fueled an ongoing debate over which model best describes human categorization. In this talk, I will describe some surprising parallels between the classic cognitive exemplar and prototype categorization models and template matching models developed in the context of visual
pattern recognition. I will then describe a series of simulations and experiments that demonstrate how externally added stimulus noise can be used to distinguish between prototype and exemplar category representations.
12/1: Bryan Bergert & Robert M. Nosofsky
Comparing Take The Best and Rational Models of Decision Making
In deciding between A and B, do we use all available information, or do we make a quick decision based on some strong feature that A has and B does not? Rational models of decision making consider all available evidence before making a decision. In the belief that the real world
contains correlations between cues that can be exploited for greater efficiency, the Take The Best (Gigerenzer & Todd 1999) model simply looks through the features in order of cue validity and makes a decision based on the first mismatching cue found.
Lee and Cummins (2002) compared optimal, deterministic versions of the Take The Best model and the Rational model on a two-alternative decision task. We generalize these models to allow nonoptimal, probabilistic behavior and show that the generalized versions provide better
quantitative fits to subject data. Because the generalized Take The Best and Rational models turn out to be formally identical in their predictions of choice probabilities, we develop a new reaction time task to distinguish between them. Finally, we apply an exemplar model of
categorization to the two-alternative decision task and compare its predictions to those of the Take The Best and Rational models.
12/8: George Chadderdon
A Neurocomputational Model of Working Memory and Task-Oriented Behavior Selection
The essential problem of volition is how to most adaptively select a response
to current stimuli from a large repertoire of possible behaviors. The mapping
between stimulus and most-adaptive response, however, is often many-to-many,
and there are times when neither external cues nor current internal drive
states are sufficient to select the behavior likely to have the most positive
consequence. It may be necessary to maintain mental representations both of
task and task parameters, holding these in working memory and responding
conditionally to both current environmental state and the maintained
representations. An example scenario of this will be presented which
involves an animal (or robot) engaging in either a delayed match-to-sample
(DMS) or a delayed non-match-to-sample (DNMS) task. Which task is cued when a
simple visual stimulus is presented will determine whether an affirming or a
negating behavior is the proper response leading to reward when a matching or
non-matching stimulus is viewed after a delay period. A large-scale
neurocomputational model will be presented that, drawing on findings from
neurophysiology, neuroimaging, and behavioral-pharmacological research,
embodies a set of explanatory hypotheses about the functioning of working
memory in prefrontal cortex (PFC) and how it may modulate response in the
described DMS/DNMS task set. One hypothesis embodied in this model of
potential clinical importance involves the role of tonic levels of dopamine in
PFC. In rat and monkey experiments, both sub- and super-normal levels of
prefrontal dopamine channel activity have been shown to impair delayed
response performance; the model to be presented will illustrate possible
mechanisms for this performance degradation. The larger intent of the model
is to explain the role of working memory in task-oriented behavior selection
and to investigate the corresponding network- and systems-level neural
mechanisms.
|
