|
The Cognitive Lunch talks will be on Wednesdays from 12:10 pm - 1:25 pm in the Psychology conference room (PY 128) located behind the main office.
Spring 2010 Schedule - 01/20 Colin Allen, Department of History and Philosophy of Science, Indiana University - Abstract
- 01/27 Christopher Harshaw, Department of Psychological & Brain Sciences, Indiana University - Abstract
- 02/03 Emmanuel Pothos, Department of Psychology, Swansea University - Abstract
- 02/10 Domenico Bertoloni Meli, Department of History and Philosophy of Science, Indiana University - Abstract
- 02/17 Winter Mason, Yahoo! Research - Abstract
- 02/24 Dana Ballard, Department of Computer Science, The University of Texas at Austin - Abstract
- 03/03 Mark Mon-Williams, Institute of Psychological Sciences, The University of Leeds - Abstract
- 03/10 Tom Wisdom, Department of Psychological and Brain Sciences, Indiana University - Abstract
- 03/24 Will Alexander, Department of Psychological and Brain Sciences, Indiana University - Abstract
- 03/31 Jared Hotaling, Department of Psychological and Brain Sciences, Indiana University - Abstract
- 04/07 Jennifer Trueblood, Cognitive Science Program, Indiana University - Abstract
- 04/14 Dan Yurovsky, Department of Psychological and Brain Sciences, Indiana University - Abstract
- 04/21 Sam Gershman, Department of Psychology and Neuroscience Institute, Princeton University - Abstract
- 04/28 George Kachergis, Department of Psychological and Brain Sciences, Indiana University - Abstract
Abstract 1/20: Colin Allen, Department of History and Philosophy of Science, Indiana University
Logic by the masses: What can be done with a million proofs?
For the better part of a decade now, at logic.tamu.edu we have been logging student attempts to solve various kinds of formal logic problems, including symbolization exercises, proofs, and countermodel construction. In this "exploratory" talk, I will (a) describe the dataset that we collected, (b) mention a use of this dataset by Landy & Goldstone, (c) describe a practical application of the data to the construction of a hint system for students, (d) explain a project we have begun, to try to understand how students use this system, and (e) appeal for ideas about other things that can be done with a million and a half student transactions.1/27: Christopher Harshaw, Department of Psychological & Brain Sciences, Indiana University
Blinking Bird Brains: The Spatiotemporal (Mis)Coordination of Attention and Orienting Affects Learning In Quail Hatchlings
A number of findings from cognitive neuroscience and psychophysics indicate the occurrence of timing-specific perceptual deficits and illusions in adult humans. For example, interaction between attention, perception, and other cognitive processes can result in “attentional blink” for the second of two target stimuli that occur in rapid succession, if the second stimulus occurs 200-500 ms after the first. We documented the occurrence and persistence of a timing-specific deficit in auditory learning in one- to three-day-old Northern bobwhite (Colinus virginianus) chicks provided with delays of 450-900 ms between their behaviors (contact vocalizations) and operant playback of a species-typical bobwhite maternal call. The deficit was found to occur only when call playback was switched back and forth semi-randomly between two speakers hidden on opposite sides of the large circular arena used for training, and not when playback was limited to a single, randomly chosen location. Detailed scoring of sessions in which chicks were provided with playback of the call at randomly chosen delays (from zero to 1250 ms) after each of their vocalizations revealed that chicks exhibit large fluctuations in their responsivity to the auditory stimulus as a function of delay that were significantly correlated with the learning data from the previous experiments. A number of further analyses provide strong evidence that the fluctuations in both chick learning and responsivity to the auditory stimulus were modulated by attentional processes and that the timing-specific deficit in learning exhibited by chicks with delays of 450-900 ms resulted from spatiotemporal conflict between (or the miscoordination of) processes of orienting and attention.2/3: Emmanuel Pothos, Department of Psychology, Swansea University
A challenge to formal models: category intuitiveness.
Are different concepts 'better' than others? We seem to have a sense in
which certain concepts are more concrete or more cohesive. For example, we
have an immediate notion of what the category of 'chairs' is about. But, try
to explain the category of 'games', and most of us will run across
difficulties. I will discuss how such differences in conceptual
intuitiveness could be measured in the laboratory and some relevant
experiments. The empirical results reveal a challenge for current
computational approaches in categorization.2/10: Domenico Bertoloni Meli, Department of History and Philosophy of Science, Indiana University
Patterns of Transformation in 17th-Century Mechanics
The 17th century (approximately from Galileo to Newton) witnessed striking transformations in the science of mechanics: whereas Renaissance authors of the previous century were primarily concerned with restoring and extending the achievements of Antiquity, 17th-century practitioners brought mechanics to radically new domains, such as the study of motion. I have recently argued that the objects or devices employed in the 17th century are a key tool for grasping such transformations in a way that reflects the contemporary practice of mathematicians and natural philosophers: levers, inclined planes, pendulums, springs, and strings were employed in a variety of fashions, both practical and theoretical, to open new areas of research and conceptualize difficult problems. In this talk I identify some patterns of transformation in this complex area, studying the creative and original methods employed by practitioners in exploring new domains. Such patterns are fertile territory for bridging historical and philosophical themes to do with research practices on the one hand, and methodological and cognitive aspects on the other.2/17: Winter Mason, Yahoo! Research
Psychological Experiments in Social Media Settings
This talk describes two research projects designed and executed on two different social media platforms: Facebook and Amazon's Mechanical Turk. In the first project, we designed a Facebook application that allowed us to measure how similar friends' attitudes are, how similar they are perceived to be, and how accurately friends know each other's attitudes. We discover that while friends are highly accurate in guessing what they believe, their actual knowledge of each other's beliefs is much more questionable. In the second project, we studied the effects of financial incentives on two tasks performed on a popular crowd-sourcing platform. We find a significant effect of incentives on quantity of output but no such effect on quality, and pose a surprising explanation for the results.2/24: Dana Ballard, Department of Computer Science, The University of Texas at Austin
Modular Reinforcement Learning as a Model of Embodied Cognition
To make progress in understanding human visuo-motor behavior, we will need to understand its basic components at an abstract level. One way to achieve such an understanding would be to create a model of a human that has a sufficient amount of complexity so as to be capable of generating such behaviors. Technological advances in VR allow significant progress to be made in this direction. Graphics models that simulate extensive human capabilities can be used as platforms from which to develop synthetic models of visuo-motor behavior. Currently such models can capture only a small portion of a full behavioral repertoire, but for the behaviors that they do model, they can describe complete visuo-motor subsystems at a useful level of detail. The value in doing so is that the body's elaborate visuo-motor structures greatly simplify the specification of the abstract behaviors that guide them. Essentially, one is faced with proposing an embodied operating system model for picking the right set of abstract behaviors at each instant. We outline one such model. Its centerpiece uses MDP reinforcement learning modules to guide behavior. (Suggested Reading)3/3: Mark Mon-Williams, Institute of Psychological Sciences, The University of Leeds
Why do we do what we do with our hands?
Human reach-to-grasp movements alter lawfully as a function of an object’s physical dimensions, its location and orientation. But what makes us grasp a particular object in a particular way? When we see an object in the world, there may be a large number of different ways to interact with that object. This large 'visuomotor space' can be constrained through affordances (perceptually available object properties defining potential uses) and the actor's intentions. But other factors can influence the action that is selected, such the comfort of the limb at the end of the movement (the end state comfort effect, ESC, Rosenbaum et al. 1990). We have found that two other biases affect interaction with a perceptually under-constrained object: hysteresis (H) and minimal forearm rotation (MR). Our experiments have shown that in addition to ESC, H and MR influence prehension behaviour in adult participants. We have also found that: (i) the influence of these biases alters over the developmental trajectory; (ii) children with neurodevelopmental disorder are biased in different ways from age matched controls; (iii) adults show different biases with their preferred versus non-preferred hand. These results argue for a conceptual separation between ‘motor’ and ‘executive’ planning’. It will be suggested that: (a) motor planning is a dynamic process of selecting actions from an extant repertoire that arises through a process of natural selection; (b) executive planning reflects abstract evaluations of possible future outcomes; (c) movement planning can be used to describe the normal synergistic relationship between motor and executive planning.3/10: Tom Wisdom, Department of Psychological and Brain Sciences, Indiana University
Social Learning and Cumulative Mutual Improvement in a Networked Group
We used a simple problem-solving game task to study imitation and
innovation in groups of participants. Guesses were composed of
multiple elements with linear and interactive effects on score, and
score feedback was provided after each of a number of rounds.
Participants were allowed to view and imitate the guesses of others
during each round, and the visibility of score information
accompanying others’ guesses was manipulated in two conditions. When
scores were not visible, social learning was impeded; participants
were less efficient in their searching of the problem space and
achieved lower performance overall. When scores were visible, higher
performance was observed, and results indicated a more equitable
sharing of productive exploration among participants within groups as
a result of selective imitation and cross-participant cumulative
mutual innovations. These results have implications for previous
models of social learning and group exploration as a social dilemma.3/24: Will Alexander, Department of Psychological and Brain Sciences, Indiana University
An impossible model of paradoxical discounting
Discounting of prospective gains is widely observed in human and animal behavior. Gains which are temporally delayed, uncertain, or require effort are less valuable than those which are available immediately, certain, or attained effortlessly. Despite its ubiquity, however, there has been relatively little success in developing a unified framework which adequately explains discounting behavior under these different domains. In this talk I present a new model of decision making, the Recursive Hyperbolic Model (RHM). The model, derived from previous models of intertemporal choice, suggests that a common mechanism, perception of reward, underlies temporal, probabilistic, and effort-based discounting. The RHM accounts for a number of inconsistent, contradictory, or paradoxical behaviors which are not fully explained by previous models of temporal discounting (e.g., exponential, average reward, & non-recursive hyperbolic discounting) or models of decision making under uncertainty (e.g., hyperbolic probability discounting, cumulative prospect theory).3/31: Jared Hotaling, Department of Psychological and Brain Sciences, Indiana University
Information Integration in Perceptual Decision Making.
In cognitive science there is a seeming paradox: On the one hand researchers studying judgment and decision making have repeatedly shown that people employ simple and often less than optimal strategies when integrating information from multiple sources. On the other hand, researchers have had great success using optimal models, often developed within a Bayesian framework, to account for information integration in fields such as categorization, memory, and perception. This apparent conflict could be due in part to different materials and designs that lead to differences in the nature of processing. Stimuli that require controlled integration of information, such as the quantitative information presented in verbal scenarios (commonly found in judgment studies), may lead to suboptimal performance. In contrast, the images often used in perceptual domains may lend themselves to automatic processing, resulting in integration that is closer to optimal. I will present research exploring these ideas within a common experimental paradigm and uniform methods of analysis. We investigated a canonical example of sub-optimal information integration from the judgment and decision making literature, the dilution effect. Surprisingly this effect was quite reliable, even with the most naturalistic stimulus presentations. However, we found evidence that manipulations meant to encourage controlled integration of information produced the largest dilution effects, while those intended to promote automatic integration yielded performance closer to optimal.4/7: Jennifer Trueblood, Cognitive Science Program, Indiana University
Explaining Order Effects: Belief-Adjustment versus Quantum Inference
One of the oldest and most reliable findings regarding human inference is that the order of evidence affects the final judgment. These order effects are non-Bayesian by nature and are difficult to explain by classical probability models. We use the empirical results of a jury decision-making task conducted by McKenzie et al. (2002) to motivate the development of a quantum inference model for order effects. The quantum model is derived from the axiomatic principles of quantum probability theory and provides a more coherent explanation of order effects than heuristic models such as the belief-adjustment model (Hogarth and Einhorn, 1992). To further test the quantum inference model, a new jury decision-making experiment is developed. This experiment extends the work of McKenzie et al., and the results are used to compare the quantum model to the belief-adjustment model. We also provide experimental evidence to suggest that the belief-adjustment model faces limitations when accounting for tasks involving extreme evidence whereas the quantum inference model does not.4/14: Dan Yurovsky, Department of Psychological and Brain Sciences, Indiana University
Linking Learning to Looking: Model Selection for Eye Movements
Measuring eye movements is rapid, easy, and unobtrusive. Because eye fixations are produced rapidly in response to stimuli, gaze patterns can be a particularly sensitive tool for measuring learning. Eye-tracking is especially important in infant research, as it is often the only available source of information about the participant’s learning. Unfortunately, inferring learning from looking in many tasks is anything but straightforward -- especially when no other measures provide convergent evidence. As a result, researchers often adopt vague - and occasionally ad hoc - metrics for linking learning to looking. But, when data is sparse and linking functions are imprecisely specified, conclusions are hotly debated and failures to replicate are common. We propose a model selection framework for choosing explicit linking hypotheses in a formal, consistent manner. We than validate the approach using two different infant learning paradigms – statistical word-learning (Smith & Yu, 2008) and cued attention (Wu & Kirkham, accepted). In each case, our framework allows us to make principled inferences about what (and how) each infant learns.4/21: Sam Gershman, Department of Psychology and Neuroscience Institute, Princeton University
Structured Reinforcement Learning
Humans can make strong inferences on the basis of little (or even no) experience. This capability is made possible by a rich repertoire of "inductive biases" that constrain the space of plausible hypotheses. Recent neural, behavioral and computational work suggests that the acquisition and exploitation of structured inductive knowledge plays an important role in human reinforcement learning. A new picture of reinforcement learning in the brain is beginning to emerge, emphasizing sophisticated interactions between simple error-driven learning mechanisms and high-level cognitive processes, in contrast to the predominant view of multiple dissociable learning systems.4/28: George Kachergis, Department of Psychological and Brain Sciences, Indiana University
Modeling the Acquisition of the Mental Lexicon
Several studies have found that adults can acquire word-referent pairings after experiencing a series of individually-ambiguous trials (i.e., trials containing multiple words and referents). To disambiguate pairings -- which many learners do with great success -- word-referent co-occurrences must be integrated across trials. I will present a variety of factors that have been found to affect human learning, such as pair frequency, pairs per trial, temporal contiguity, and contextual diversity (how many pairs a given pair appears with during training). Using an associative learning framework, I will discuss the attentional and memory mechanisms that account for these data, including entropy-based and Bayesian models.
Previous Cognitive Lunch
|
