Sixteenth Annual Summer Interdisciplinary Conference

Authors, Titles, Abstracts

Listing by speaker

SpeakerAnderson, John
Author 1Anderson, John
Carnegie Mellon
TitleThe 3-State Structure of Memory and the Environment
AbstractItems that we encounter in our environments appear as if they are in one of three states: Lost, Background, or Current. For instance, chimpanzees in the Stevens, Marewski, Schooler, and Gilby (2016) data set had either had not been seen for years, were irregularly seen, or were regularly seen. They make appear to make discrete changes in these states. Similar patterns appear in much larger databases, for instance strings used in Twitter messages. Human memory acts as if it were making inferences about which state an item is in, and adjusting its availability accordingly, accounting for discrete changes in retrieval times.

SpeakerAtmanspacher, Harald
Author 1Atmanspacher, Harald
Collegium Helveticum, ETH and University Zurich
TitleNon-Commutative Operations in Psychology
AbstractIt will be argued that measurements whose sequence makes a difference for their results should be the rule rather than the exception in psychology. The mathematical formalism that can be applied to model such operations is based on non-commutative algebras of operations, as they are widely used in quantum theory. The basic structure of such algebras will be outlined and illustrated with one or two examples.

SpeakerAwh, Edward
Author 1Awh, Edward
University of Chicago
Author 2Sutterer, David
University of Chicago
Author 3Foster, Joshua
University of Chicago
TitleTracking the timing and content of retrieval from long term memory with rhythmic brain activity
AbstractRecent work has shown that the spatial distribution of alpha oscillations on the scalp tracks spatial representations in working memory and the locus of covert spatial attention. Here, we exploit the same signal to track the timing and content of retrieval from long term memory. We found that spatially-selective activity in the alpha band emerged approximately 500 ms after subjects were cued to retrieve a position that had been previously associated with a shape cue. This spatial representation was temporally coincident with the "parietal old/new" event-related potential that has previously been shown to index successful recollective events. Thus, we provide a powerful new approach for tracking both content and timing during retrieval from long term memory.

SpeakerBorner, Katy
Author 1Borner, Katy
TitleData Visualization Literacy: Definitions, Measures, Means to Advance
AbstractThis talk present the results of a recent study that examined the “data visualization literacy” of over 900 youth and adult visitors across six U.S. science museums. Results show that: a very high proportion of the population, both adult and youth, cannot interpret data visualizations beyond very basic reference systems; construction of complex visualizations led to more accurate interpretation than deconstruction; and individuals are willing to spend time attempting to make meaning in representations depending on their personal interest in the topic. I then discuss a theoretically grounded and practically useful visualization framework that was developed to empower the broadest spectrum of users to read and make data visualizations that are useful and meaningful to them. The visualization framework was used to conduct the aforementioned study and is is used to develop plug-and-play macroscope tools that improve the (network) data visualization literacy of researchers, practitioners, IVMOOC students, museum visitors, and others.

SpeakerBurns, Devin
Author 1Burns, Devin
Missouri University of Science & Technology
TitleDimension specific sequential effects in the Garner Paradigm
AbstractIn the Garner paradigm (and other tests of selective attention), stimuli can vary either along a “relevant” dimension which fully indicates the correct response or an “irrelevant dimension” which carries zero information about the correct response. Strong sequential effects have been shown to occur depending on which of these dimensions changes or repeats from one trial to the next, as modeled by Little, Wang, and Nosofsky (2016). This work examines these effects in a three dimensional extension of the Garner paradigm (Burns, 2016) where one dimension (orientation) is separable from an integral pair of dimensions (saturation and hue). When the irrelevant dimension is separable from the relevant dimension, it's variation has no contribution to sequential effects, as expected. In this case, trials where the relevant dimension repeats should be faster than those when it changes, which occurs when orientation is being ignored in favor of either saturation or hue. However, the opposite result occurs when orientation is attended to, with trials where this dimension changes being faster than repeats! Possible causes and implications will be discussed.

SpeakerChen, Sharon
Author 1Chen, Sharon
Syracuse University
Author 2Criss, Amy
Syracuse University
TitleThe Source of List Strength Effect within REM: The Level of Competition
AbstractTo study episodic memory, we study interference. With a list strength paradigm, we study how such interference is affected by how well stimuli are encoded, and what kind of list they are in. A stimulus can be weak or strong, and it can be in a pure list, composed of all weak or all strong stimuli, or a mixed list, composed of both weak and strong stimuli. A list strength effect (LSE) refers to the interaction between stimulus strength and list type. In free recall, where the cue used at test is only context, we have consistently observed a positive LSE. Yet, in cued recall, where the cue is made of item, we have consistently observed a null LSE. Thus, we attributed the source of LSE to the type of cue used at retrieval. Based on REM, this framing is misleading. It is not the type of cue (context or item) that is critical, but the level of competition. In the citations, these two factors are confounded. Therefore, in this study, we manipulated the level of competition (between-subject) and the type of cue (within-subject) simultaneously in the list strength paradigm. Data shows LSE was determined by the level of competition, not the type of cue used to probe memory at test. Fitting REM with data confirms this statement. Nevertheless, context and item cue affected recall performance differently. Implications for models of memory will be discussed.

SpeakerDixon, Peter
Author 1Dixon, Peter
University of Alberta
TitleAssessing comprehension and mind wandering
AbstractA common technique in research on mind wandering is to periodically ask subjects to self report their mental state: Were they on task or were they mind wandering? Such reports often predict performance: For example, text memory is generally worse when readers say they were mind wandering. However, this relationship doesn't imply that self reports are perfectly valid; it is still possible that any given variable only affects self report and not mind wandering (or vice versa). In order to further understand the relationship between self reports and mind wandering, I describe how text memory is related to self reports across manipulations of text interest, reading interruptions, and visual distractions. I argue that mental state reports are best understood as an indirect inference rather than as a direct, metacognitive index.

SpeakerFourny, Ghislain
Author 1Fourny, Ghislain
ETH Zürich
Author 2Reiche, Stéphane
Mines Paris
Author 3Dupuy, Jean-Pierre
Stanford University
TitlePerfect Prediction Equilibrium: Non-Nashian dynamic games
AbstractToday, the most established paradigm of game theory is the Nash equilibrium. It predicts what people do based on the assumption that people are "rational", which makes them predictable to some extent. Many people think the Nash way. Yet, the Nash paradigm doesn't account for all our actions, in particular, it cannot explain why we keep our promises with no further incentive. Is it irrational to be honest? There is a mind-blowing problem that is well known especially in the US: Newcomb's problem. It points out that people have two different mindsets when it comes to predicting people's behavior. According to the Nash paradigm, it is rational (dominant strategy) to take two boxes ($1000). However, while most game theoreticians, physicists, etc, do take two boxes, research shows that 75% of the people take only one box, and get a higher payoff ($1000000). Yet no general game theory framework to date accounts for this observed behavior, or for promise-keeping in non-cooperative games. This means that there is a large territory to explore out there that we could stamp as Non-Nashian Game Theory, as an analogy to Non-Euclidian Geometry. We designed the Perfect Prediction Equilibrium as an equilibrium for dynamic games that accounts for this alternate mindset. In short, with this equilibrium, people integrate in their reasoning that they have such accurate logical and predictive skills, that a prediction is counterfactually dependent on what it predicts: The world is fully transparent. The Perfect Prediction Equilibrium (PPE) then emerges as the solution to a fixpoint problem that can be expressed in first-order logics. In this talk, I will introduce the PPE, as well as put it in perspective with other (growing) Non-Nashian literature for the normal form (Douglas Hofstadter's Superrationality as well Joseph Halpern's CCBR) or extensive form (Richard Shiffrin's locally rational decision making).

SpeakerFrench, Bob
Author 1French, Bob
LEAD-CNRS, U. of Burgundy, France
Author 2Mareschal, Denis
Birkbeck University of London, U.K.
TitleTRACX2: a connectionist autoencoder using graded chunks to model infant visual statistical learning
AbstractEven newborn infants are able to extract structure from a stream of sensory inputs; yet how this is achieved remains largely a mystery. We present a connectionist autoencoder model, TRACX2, a successor of TRACX (French et al. 2011) that learns to extract sequence structure by gradually constructing chunks, storing these chunks in a distributed manner across its synaptic weights and recognizing these chunks when they re-occur in the input stream. In TRACX2, compared to the original TRACX, chunks are graded rather than all-or-nothing in nature. This reflects the (rather obvious) fact that chunks in the real world are graded. There is a smooth continuum between words like "cupboard" (who thinks "cup" and "board" when they read this word?) to words like "smartphone" in which everyone still hears "smart" and "phone". As chunks are learned their component parts become more and more tightly bound together. TRACX2 successfully models the data from five experiments from the infant visual statistical learning literature, including tasks involving forward and backward transitional probabilities, low-salience embedded chunk items, part-sequences and illusory items. It also models as well, or better than, TRACX data from a variety of auditory statistical learning experiments. The model also captures performance differences across ages through the tuning of a single learning rate parameter. These results suggest that infant statistical learning is underpinned by the same domain-general learning mechanism that operates in auditory statistical learning and, potentially, in adult artificial grammar learning.

SpeakerGoldstone, Robert
Author 1Goldstone, Robert
Indiana University
Author 2Mills, Brian
University of Florida
TitleBias and Learning in Major League Baseball Umpires’ Perceptual Judgments
AbstractMajor League Baseball home plate umpires have collectively made millions of professional pitch calls, and these calls can be compared to trajectory information recorded since 2008 for each pitch using tracking technology that is accurate to within 1 MPH and 1 inch. Furthermore, pitch, umpire, and game data are publicly available and relatively easily scraped using modern analysis tools. Using this data, we are interested in characterizing how umpires' perceptual judgments are influenced by situational factors and their own experience making calls. We fit a parametric model to account for variation in judgment policies in terms of a strike zone's horizontal and vertical center, shape, and sharpness, and the umpire's guessing probability and bias.

SpeakerGorea, Andrei
Author 1Gorea, Andrei
CNRS and Université Paris Descartes
Author 2Granjon, Lionel
CNRS and Université Paris Descartes
Author 3Sagi, Dov
Weizmann Institute
AbstractHow much do humans know about the outcome of their own actions? Participants performed a speeded pointing task, moving their index finger to a location briefly marked by a visual target, with and without visual feedback. They then estimated (E) their landing location (L) using the same finger with no time limit. Targets (T) were randomly flashed in one of three screen regions, straight ahead and 45° to the left and to the right of the central fixation cross. Overall, L and E precisions (1/SD) are very similar and are highly correlated across participants. This similarity argues against a Bayesian account of the results inasmuch as L and E distributions are taken to represent the prior and the posterior. L and E locations are systematically but idiosyncratically shifted away from T and L locations, respectively, suggesting participant specific errors in the sensorimotor transformation from the visual localization of T to the joint-based motor command. Trial-by-trial L-E correlations reveal an E bias away from L and toward the T of 20% and 36% in the presence and absence of visual feedback, respectively. The data are well accounted for by a model rooted in the general principles of the forward internal models. It posits that participants do not know the outcome of their pointing movements and, when asked to introspect, refer to their planned rather than actual landing locations. The end product of both action (landing) and introspection (estimation) is the noisy planned action outcome to which motor execution and introspection add independent, participant specific noises.

SpeakerGrünwald, Peter
Author 1Grünwald, Peter
CWI and Leiden University
TitleSafe Hypothesis Testing
AbstractPrompted by the "reproducibility crisis" in science, the 80-year old debate on how to test a hypothesis is back with a vengeance. P-value based tests are under fire, Bayesian and confidence-based approaches, while mutually incompatible, are on the rise, and information-theoretic approaches linger in the background. Here we propose a new method that, we argue, by and large combines the best of these four worlds. This safe testing method first and foremost has a gambling interpretation - evidence against a hypothesis is measured by the amount of money gained when betting under odds that would be fair if that hypothesis were true. Safe testing allows for 'optional continuation': if a test result is promising but inclusive (say, in terms of p-values, we observe p = 0.07) then, if indeed we work with p-values, we cannot simply decide to gather a few more data points; this invalidates our frequentist error guarantees. With safe testing though, guarantees are preserved. It has very often been claimed (most recently by Rouder, "Optional Stopping: No Problem for Bayesians") that Bayesian approaches have no issues with optional continuation. We show that, in general, this is quite simply false - it just happens to be true for the Jeffreys-Bayesian t-test that Rouder considers. This JB t-test is, in fact, a safe test, and we characterize exactly those Bayesian tests which are also safe tests. We also show how to incorporate prior knowledge into safe tests, and how they relate to MDL.

SpeakerGuest, Olivia
Author 1Guest, Olivia
University College London
Author 2Love, Bradley
University College London
TitleDeep Networks as Models of Human and Animal Categorization
AbstractConvolutional neural networks (CNNs) trained as classifiers learn by associating visual inputs (e.g., photographs) with labels (e.g., "crow'', "dog", "car"), rivalling humans in object recognition tasks. The similarity spaces created at various network layers allow us to draw parallels with the human brain's neural coding schemes. At earlier layers, networks display similarity spaces that reflect high-level categories found in the input space, e.g., lions and tigers are more similar to one another than to mopeds. At advanced layers, similarity structure tends to break down such that representations of different object categories become orthogonal. Can these networks also shed light on how non-human animals categorize? CNNs can be used to determine at what level of representation animals are coding similarities between images. Are animals learning regularities at low levels, close to the pixels, or are they extracting abstract shape features? We address these questions using data from pigeons which have been shown to excel at classifying visual stimuli. For example, pigeons trained to discriminate between medical images of normal and cancerous breast tissue generalized to novel stimuli, attaining human expert-level accuracy. However, knowledge transfer was limited, e.g., they did not generalise between magnification levels or to greyscale images — although performance improved with additional training on greyscale images. Can CNNs explain such patterns of performance? Which network layer best captures how pigeons categorize? We consider the broader implications of the answers to these questions for how humans and non-human animals categorize.

SpeakerHalpern, Joseph
Author 1Halpern, Joseph
Author 2Kleiman-Weiner, Max
TitleTowards a Formal Definition of Blameworthiness, Intention, and Moral Responsibility
AbstractWe provide formal definitions of degree of blameworthiness and intention relative to an epistemic state (a probability over causal models and a utility function on outcomes). These, together with a definition of actual causality, provide the key ingredients for moral responsibility judgments. We show that these definitions give insight into commonsense intuitions in a variety of puzzling cases from the literature.

SpeakerHalpern, Joseph
Author 1Halpern, Joseph
Author 2Pass, Rafael
Author 3Seeman, Lior
TitleDecision theory with resource-bounded agents
AbstractThere have been two major lines of research aimed at capturing resource-bounded players in game theory. The first, initiated by Rubinstein, charges an agent for doing costly computation; the second, initiated by Neyman does not charge for computation, but limits the computation that agents can do, typically by modeling agents as finite automata. We review recent work on applying both approaches in the context of decision theory. For the first approach, we take the objects of choice in a decision problem to be Turing machines, and charge players for the ``complexity'' of the Turing machine chosen (e.g., its running time). This approach can be used to explain well-known phenomena like first-impression-matters biases (i.e., people tend to put more weight on evidence they hear early on) and belief polarization (two people with different prior beliefs, hearing the same evidence, can end up with diametrically opposed conclusions) as the outcomes of quite rational decisions. For the second approach, we model people as finite automata, and provide a simple algorithm that, on a problem that captures a number of settings of interest, provably performs optimally as the number of states in the automaton increases. Perhaps more importantly, it seems to capture a number of features of human behavior, as observed in experiments.

SpeakerHamilton, Larry
Author 1Hamilton, Larry
University of New Hampshire
TitleIdeological Asymmetry of Science Rejection
AbstractIdeology-linked distrust of scientists, and rejection of scientific findings, have been observed on both left and right. Theories predict, and some experiments confirm, that similar cultural-cognition processes of identity protection, biased assimilation, motivated reasoning and elite cues can occur under diverse ideological orientations. These theories and experiments might lead us to expect a degree of left-right symmetry in science rejection -- but that expectation is rudely contradicted by current events, and also by general-public surveys. In the U.S., conservative elites and citizens have been much more likely than liberals to reject scientific findings across a range of central topics where scientists themselves overwhelmingly agree, such as evolution, climate change, vaccinations, and the age of the Earth. Moreover, surveys find disproportionate conservative distrust of scientists extending to other topics where commentators predicted the opposite pattern (nuclear power and genetically modified organisms), and also to emergent topics not previously connected to political, economic or cultural beliefs (such as the Zika virus). Many different surveys asking different questions have found common ideological gradients in distrust of science, a pattern incompatible with the usual cognitive and sociological explanations. Detailed analyses of these data suggest what hypotheses fail, and what might succeed, in explaining the ideological asymmetry of science rejection.

SpeakerHanson, Andrew
Author 1Hanson, Andrew
Indiana University
TitleThe 4D Room
AbstractHow does one design and exploit interactive environments that enable the comprehension of a higher dimensional space? How does one determine whether a higher dimensional space is "understood?" Here we focus on the transition from living in 3D space to exploring a virtual 4D space, as even that small step seems to embody substantial effort. In earlier work, we have discussed the design and implementation of "4Dice," an iPhone App that uses the rotational behavior of a 3D die as a tutorial, and then provides a full-featured extension to the rotational behavior of a 4D die, with the intention of exposing every possible analogy to 3D space to aid in the absorption of the new properties of 4D space. Here we will describe the extension of "4Dice" to "4DRoom," a new iPhone App that moves the viewer to the center of 3D and 4D die, looking outward at the walls. One can start inside a 3D die, effectively a 3D room environment, the potential starting point for a 3D maze exploration. The second stage of the "4DRoom" application, obviously, moves the viewer to the interior of a 4D die, which embodies all the correct properties of a four-dimensional room, and provides the environment in which the analogies between the familiar 3D room and the strange 4D room can be explored in manifestly correct and rigorous geometric detail. Our 4D room application has the additional feature of supporting an optional Virtual Reality style control, with changing views accessed by rotating the hand-held device in space, as though moving one's head around in a virtual reality game, or in a New York Times VR Google Cardboard environment. We will conclude with a discussion of the issues involved in designing 2D, 3D, and 4D mazes based on sequences of room-like environments such as those supported by "4DRoom," and the question of whether one can conclude that the cognitive process of "understanding space" can be demonstrated by measuring the behavior of mice and (wo)men as they make their way through such mazes. The relevant applications, 4Dice and 4DRoom, are available free for iPhone and iPad environments on the iTunes App store: see and .

SpeakerHoward, Marc
Author 1Howard, Marc
Boston University
TitleToward a general model of episodic memory
AbstractOver the past several decades, mathematical models of episodic memory tasks have clustered around two tasks. Recognition models have developed largely independently of recall models. Recall models have hypothesized that recall is moderated by a gradually-changing state of context. A subset of recall models hypothesize further that episodic memory is accompanied by recovery of a prior state of temporal context. In recent years, neurophysiological evidence has provided dramatic evidence for both of these conjectures. I will discuss the implications of these developments for a mathematical model of recognition memory based on a scale-invariant representation of temporal context.

SpeakerKhrennikov, Andrei
Author 1Khrennikov, Andrei
Linnaeus University
TitleQuantum-like model for social laser
AbstractWe discuss applications of quantum formalism to cognition and behavioural economics. Starting with consideration of conjunction and disjunction effects, we proceed to consider applications to decision making and common knowledge. Finally, macro-processes in society are compared to effects of excitation amplification as used in laser physics.

SpeakerLewandowsky, Stephan
Author 1Lewandowsky, Stephan
University of Bristol
TitleWhen denial of the undeniable and logical incoherence are politically rational: An anatomy of climate change denial in an era of 'post-truth' politics"
AbstractAlthough the relevant scientific community long ago settled on the conclusion that human economic activities are causing climate change through the emission of greenhouse gases, a small but vocal number of dissenters remains unswayed by the evidence. I examine the cognitive and motivational factors that underlie the rejection of scientific evidence, and I illustrate the techniques by which contrarians seek to shape public debate and mislead the public. Many of these techniques involve tacit or overt appeals to conspiracy theories, and much contrarian discourse exhibits features of conspiracist ideation. However, at a higher political level of abstraction, denial of climate change is found to be an entirely rational and politically successful activity.

SpeakerLittle, Daniel R.
Author 1Cheng, Xue Jun
The University of Melbourne
Author 2McCarthy, Callum
The University of Melbourne
Author 3Wang, Tony S. L.
Brown University
Author 4Palmeri, Thomas J.
Vanderbuilt University
Author 5Little, Daniel R.
The University of Melbourne
TitleComposite faces are not (necessarily) processed coactively
AbstractUpright faces are thought to be processed more "holistically" than inverted faces. In the composite face paradigm, holistic processing is inferred from decreased recognition performance for upright and aligned composite faces compared to inverted and misaligned faces. However, the composite face task does not necessarily address the nature of holism in the way that the term is sometimes defined computationally. In a categorization task, we use the logical-rule models (Fifić, Little, & Nosofsky, 2010) and Systems Factorial Technology (Townsend & Nozawa, 1995) to examine whether composite faces are processed through pooling top and bottom face halves into a single processing channel (i.e., coactive processing) which is one mechanistic definition of holistic processing. By specifically operationalising holistic processing as the pooling of features into a single decision process in our task, we are able to distinguish it from other processing models that may underlie composite face processing. For instance, a failure of selective attention (another common definition of holistic processing) might result even when top and bottom components of composite faces are processed in serial or in parallel. Our results show that performance is best explained by a mixture of serial and parallel processing architectures across all four upright and inverted, aligned and misaligned face conditions. The results indicate multi-channel, featural processing of composite faces which is inconsistent with the notion of coactivity.

SpeakerLove, Bradley
Author 1Love, Bradley
University College London
TitleExploration with Objective and Subjective Rewards
AbstractIn uncertain environments, effective decision makers balance exploiting options that are currently preferred against exploring alternative options that may prove superior. For example, a honeybee foraging for nectar must decide whether to continue exploiting the current patch or move to a new location. When the relative reward of options changes over time, humans explore in a normatively correct fashion, exploring more often when they are uncertain about the relative value of competing options. The prevailing view is that exploratory behaviour is uncertainty minimising, capacity-limited, and frontally mediated. However, rewards in these laboratory studies were objective (e.g., monetary payoff), whereas many real-world decision environments involve subjective evaluations of reward (e.g., satisfaction with food choice). In such cases, rather than choices following preferences, preferences may follow choices. With subjective rewards, rather than minimise uncertainty, people may seek to maximise coherency between their preferences and behaviour. If so, increasing coherency would lessen the tendency to explore while uncertainty increases, contrary to previous findings. Indeed, a study of 280,000 individuals in supermarkets over several years found just this pattern of exploration. Consumers' patterns of exploratory choice ran counter to normative models for objective rewards – the longer the exploitation streak for a product, the less likely were people to explore an alternative. I'll present a model that captures exploratory behaviour with both objective and subjective rewards by characterising reward signals as multidimensional. The model may help explain superficially different behaviours, such as political polarisation and confirmatory biases.

SpeakerMar, Raymond
Author 1Mar, Raymond
York University
TitleThe SPaCEN Framework for Studying how Stories Relate to Social Cognition
AbstractA great deal of research evidence points to an association between exposure to fictional narratives (e.g., books, movies, TV shows) and social cognitive processes such as mental-inference ability and empathy. However, these disparate studies have yet to be synthesized into a formal framework that makes formalizes the exact nature of how, when, and why engagement with fiction might help to promote social cognition. I will present the early workings of such a model, entitled Social Processes and Content Entrained by Narrative (SPaCEN). It argues that there are two main routes through which stories could promote social cognition, (1) through frequent engagement of social-cognitive processes or (2) the presentation of explicit content about the social world. These two possibilities are not mutually exclusive and discussion of work in this area has occasionally confused the two. I hope to outline more clearly the implications of both the process and content accounts and review the supporting evidence for each. One thing that the SPaCEN model highlights is that although a great deal of research on this topic currently exists, there remain numerous lacunae that remain unanswered and often unaddressed.

SpeakerMunro, Paul
Author 1Munro, Paul
University of Pittsburgh
TitleEmergence of a teacher signal from a self-supervised learning procedure
AbstractMany activity dependent learning rules have a positive and negative term that operate oppositionally. Here, a hypothetical system is described for a neuron that is partitioned into two subunits (T and L) which integrate their inputs independently. The synapses of both subunits are modified according to a learning rule that resembles the form of the Delta Rule; e.g. the Perceptron Learning Rule. Here, both subunits have a common positive term (analogous to a training signal) that is a nonlinear function of the response of one of the subunits (T), suggesting a common biological mechanism. Thus, the learning rule for T is unsupervised. The rule also incorporates a modifiable threshold and is a version of the BCM rule (an unsupervised learning rule that maximizes selectivity across a pattern space). Thus, subunit T develops a response profile that is highly selective over its stimulus environment. The positive term plays the role of a training signal for L. The result is that the L subunit develops a response profile that predicts T based on the inputs to L. This gives a biologically plausible and mathematically parsimonious account for the development of a teacher signal. Biologically, the two subunits could correspond to morphological features of a neuron; for example, in a pyramidal cell, the major dendritic trees could be the subunits. Finally a toy simulation is presented demonstrating this learning rule as a possible mechanism for the role of language in the shaping of concepts.

SpeakerNeufeld, Jim
Author 1Neufeld, Jim
University of Western Ontario
Author 2Taylor, Reggie
University of Western Ontario
Author 3Cutler, Collleen
University of Western Ontario
Author 4Theberge, Jean
University of Western Ontario
Author 5Densmore, Maria
University of Western Ontario
Author 6Williamson, Peter
University of Western Ontario
TitleAnalytical Cognitive Modeling is a Linchpin between Levels of Neuroimaging in Schizophrenia
AbstractContemporary forms of functional neuroimaging include vascular, Blood-Oxygen-Dependent (BOLD) MRI (fMRI), Magnet Resonance Spectroscopy (fMRS), electromagnetoencephalography (MEG), and Evoked response Potential (ERP). Although differing in spatial and temporal resolution of neuroimaging signals, alternate forms of neuroimaging can yield complementing information about cognition-related neuro-activation and neuro-connectivity (co-activation). A point of contact for the alternate forms of neuroimaging is that of a common cognitive challenge. Ascertaining the stability of cognitive performance across different forms of neuroimaging, occasions of measurement, settings of measurement, etc., arguably demands valid analytical monitoring of the presumed MRI transcending cognition. To serve as a communalizing agent, invariance of the “f” of fMRI, etc., across the levels of functional neuroimaging should be quantitatively established. The described strategy is demonstrated in a functional cognitive neuroscience study of symptom-significant Stroop-Task performance in schizophrenia. Functional MRS (addressing functional ACC glutamatergic mechanisms) was undertaken among schizophrenia, and psychiatric, and healthy control participants, followed 1.5 hours later by an fMRI study (addressing neuro-circuitry, through BOLD-signal seed-voxel time-series covariance) all at 7.0 Tesla, and using the same Stroop paradigm throughout. Parametric and non-parametric stochastic modeling of latency distributions indicated consistency of model architecture and parameterized schizophrenia abnormalities. Convergent support for the referent process model includes strategic implementation of Euclid's lemma in addressing very short stimulus encoding times, and robustness with mixture-model extensions. It is noted that targeted events in event-related neuroimaging are expressly cognitive processes, whose intra-trial time trajectories must be delineated analytically. Signal-processing methodology commensurately increasing the temporal resolution of neuroimaging data is described.

SpeakerOberauer, Klaus
Author 1Oberauer, Klaus
University of Zurich
Author 2Bartsch, Lea
University of Zurich
TitleWorking memory for items and bindings
AbstractWhat is limited by the limited capacity of working memory? One of us has argued, based on individual-differences studies, that working memory capacity reflects the ability to maintain temporary bindings between representations (Oberauer, 2005, JEP:G). This hypothesis motivated a series of experiments aimed to separately measure memory for individual items and for bindings between them. Participants remembered a variable number of word-picture pairs. In most experiments, each word and each picture was used only once in the experiment. At test they were given one element of a pair and had to choose the other element from four options: the correct element, another element from the present trial, another element from a previous trial, and a new stimulus. With increasing number of pairs to be remembered, participants made more confusions between pairs of the current trial, but they rarely chose a stimulus from a previous trial, and hardly ever chose a new stimulus. Hence, memory load primarily decreased memory for bindings. At the same time, these experiments revealed a number of unexpected findings. First, increasing memory load beyond 4 pairs had only a negligible effect on accuracy, contrary to what would be expected from a limited-capacity working memory. Second, memory was immune to proactive interference, contrary to what would be expected if long-term memory contributed substantially to performance. Third, a distractor processing task (mental arithmetic) during the retention interval impaired memory for small but not larger set sizes. Together, these results challenge all contemporary theories of (working) memory.

SpeakerPopov, Vencislav
Author 1Popov, Vencislav
Carnegie Mellon University
Author 2Hristova, Penka
New Bulgarian University
TitleThe Relational Luring Effect: Retrieval of Relational Information during Associative Recognition
AbstractI will argue that semantic relations (e.g., works in: nurse hospital) have abstract independent representations in long-term memory and that the same representation is accessed by all exemplars of a specific relation. I will present evidence from two associative recognition experiments that uncovered a novel relational luring effect (RLE) in recognition memory. Participants studied word pairs, and then discriminated between intact (old) pairs and recombined lures. In the first experiment participants responded more slowly to lures that were relationally similar (table cloth) to studied pairs (floor carpet), in contrast to relationally dissimilar lures (pipe water). Experiment 2 extended the RLE by showing a continuous effect of relational lure strength on both RTs, false alarms and hits. It employed a continuous pair recognition task, where each recombined lure or target could be preceded by 0, 1, 2, 3 or 4 different exemplars of the same relation. RTs and false alarms increased linearly with the number of different previously seen relationally similar pairs. Moreover, more typical exemplars of a given relation lead to a stronger RLE. Finally, hits for intact pairs also rose with the number of previously studied different relational instances. These results suggest that semantic relations exist as independent representations in LTM, and that during associative recognition these representations can be a spurious source of familiarity. I will discuss the implications of the RLE for current models of semantic and episodic memory, unitization in associative recognition, analogical reasoning and retrieval, as well as constructive memory research.

SpeakerPopov, Vencislav
Author 1Popov, Vencislav
Carnegie Mellon University
Author 2Reder, Lynne
Carnegie Mellon University
TitleRepetition improves memory by strengthening existing traces: Evidence from paired-associate learning under midazolam
AbstractThe nature of the memory trace that supports recognition and recall has been a fundamental issue in memory theories. While cumulative-strength theories suggest that a single trace is strengthened after each repetition of an item, multiple-trace theories argue that repetitions create novel mnemonic traces. Here, we examined eventual cued-recall performance of paired-associates that were initially studied, tested and re-studied repeatedly without any successful recall before an injection of midazolam, a benzodiazepine that prevents the storage of novel associations. Their eventual recall performance was contrasted with other pairs that were only studied, tested and re-studied after the injection of midazolam. For both types of pairs that were included in the contrast there was a study and a failed cued-recall test after the midazolam injection. Only the pairs that had been studied prior to the midazolam injection benefitted from restudy. These results suggest that memory traces existed for the pairs that had been studied prior to the injection but that the activation or strength of those memory traces in previous tests was under the retrieval threshold. Those traces could be strengthened during the restudy phase even under midazolam. In contrast, word pairs that had not been studied prior to the injection were not afforded an opportunity to be bound in long-term memory (LTM) because midazolam prevents the formation of new associations. These data support the cumulative-strength explanations of repetition on memory. Finally, we present a simple computational model that accounts for these effects.

SpeakerRatcliff, Roger
Author 1Ratcliff, Roger
Ohio State University
Author 2Mckoon, Gail
Ohio State University
TitleModeling Decision Processes on a Continuous Scale
AbstractWe present a model for perceptual decision making for stimuli and responses in continuous space on lines, circles, and planes. Applications use a range of stimulus types, including perceptual, symbolic, dynamic, and static. However in this talk we will present applciations to two numeracy tasks. In one, a two-digit number is presented and the participant has to move their finger to a matching location on a number line, in the other an array of dots is presented and the participant is to move their finger to a matching location on an arc. The models are diffusion processes on lines and planes. In the models, evidence from a stimulus drives the noisy decision process which accumulates evidence over time to a criterion at which point a response is initiated. Noise is represented as a continuous Gaussian process or Gaussian random field. The model produces predictions for the full distributions of response times and choice probabilities and fits to data for choice probability, RT distributions, and choice proportion and RT across the stimulus space are presented.

SpeakerRatcliff, Roger
Author 1Ratcliff, Roger
Ohio State University
Author 2Mckoon, Gail
Ohio State University
TitleModeling Decision Processes on a Continuous Scale
AbstractWe present a model for perceptual decision making for stimuli and responses in continuous space on lines, circles, and planes. Applications use a range of stimulus types, including perceptual, symbolic, dynamic, and static. However in this talk we will present applciations to two numeracy tasks. In one, a two-digit number is presented and the participant has to move their finger to a matching location on a number line, in the other an array of dots is presented and the participant is to move their finger to a matching location on an arc. The models are diffusion processes on lines and planes. In the models, evidence from a stimulus drives the noisy decision process which accumulates evidence over time to a criterion at which point a response is initiated. Noise is represented as a continuous Gaussian process or Gaussian random field. The model produces predictions for the full distributions of response times and choice probabilities and fits to data for choice probability, RT distributions, and choice proportion and RT across the stimulus space are presented.

SpeakerReder, Lynne
Author 1Reder, Lynne
Carnegie Mellon
Author 2Popov, Vencislav
Carnegie Mellon
TitleWorking Memory Needs to be Re-conceptualized
AbstractThree experiments provide evidence that working memory resources are consumed as an inverse function of the familiarity of the chunks or elements being processed: In Experiment 1, Chinese characters are differentially familiarized in a visual search task. Ability to learn arbitrary pairings of these characters and recall an associated English word is easier when the constituent characters are more familiar; Furthermore, performance in an n-back is superior for blocks that use more familiar Chinese characters as stimuli, especially as the working memory load increases. Experiment 2 replicated the N-back task using pictures of “Fribbles” whose features can be more carefully controlled than Chinese characters, obviating the need to correct for encoding efficiency. The role of familiarization on consumption of working memory resources was self-evident. Experiment 3 demonstrated the effects of experimental familiarization on higher-level cognition with an algebra task that required differential use of working memory resources. In the task, participants had to solve algebraic equations with different complexity (1 vs 2 steps), and on half of the trials they had to substitute Chinese characters for associated digits. On each trial, we assigned numeric values to Chinese characters, previously familiarized in a different task, as the variables to be substituted into the equation to be solved. Performance was better when characters were highly familiar, and the effect increased as the complexity of the equation increased and when the equation required character substitution. The results strongly support the view that consumption of working memory resources depends on the familiarity of the elements to be processed.

SpeakerRieskamp, Jörg
Author 1Rieskamp, Jörg
Department of Psychology, University of Basel
Author 2Olschewski, Sebastian
Department of Psychology, University of Basel
Author 3Scheibehenne, Benjamin
Business School, University of Geneva
TitlePreference shifts or more errors: How increased cognitive load changes decision making
AbstractHow do people make preferential choices in situations where their cognitive capacities are limited? In the literature, many studies link the manipulation of cognitive resources to qualitative changes in preferences. However, there is a widely overlooked alternative hypothesis: Namely that a reduction of cognitive capacities leads to an increase in choice error and hence less reliable preferences. We developed a mathematical model and a hierarchical Bayesian estimation to test to what extent a reduction in cognitive capacities leads to a shift in preferences or an increase in choice error. Using a within-subject n-back task to manipulate cognitive load, we conducted three experiments across different choice domains, including risky choice, temporal discounting, and strategic interaction (ultimatum game). Across all three domains results show that a reduction in cognitive capacity credibly affected participants' level of choice error rather than their respective preferences. These results hold on an individual and on a group level. In sum, our approach and the mathematical model we used provides, contrary to past work, a rigorous test of how reduced cognitive capacity affects people's decision making behavior.

SpeakerShiffrin, Richard
Author 1Shiffrin, Richard
Indiana University
TitleGame Theory: Rational Decision Equibrium
AbstractIn this Game Theory talk, I present an algorithm that converges on an optimal unique set of decisions by two purely selfish rational agents making sequential decisions in a large decision tree with payoffs for both agents at every leaf. Both agents have full knowledge of the decision tree and the fact that other agent does as well, and that the other agent is also selfish and a rational decision maker. The key assumption is that the two agents think similarly, and each can anticipate the reasoning of the other agent. The algorithm implements a recursive set of levels of reasoning, thus traversing each node in the decision tree multiple times, like a Chinese Rings puzzle. The premises are akin to those presented in another ASIC talk by Fourny, called ‘Perfect Prediction Equilibrium', or PPE. However the two algorithms differ in critical ways, chiefly because PPE goes only one level deep in reasoning. I will make a case for my algorithm and against PPE.

SpeakerShiffrin, Richard
Author 1Shiffrin, Richard
Indiana University
Author 2Chandramouli, Suyog
Indiana University
TitleBayesian Assessment of Validity
AbstractThe present ‘crisis of reproducibility' refers to the publication of many results that are not replicable or reproducible, and points to many reasons, such as ‘file drawer effects', ‘publication bias', ‘inappropriate statistical criteria', ‘the wrong incentive structure', and ‘a variety of largely unconscious biases'. We point out that the real issue is validity, not reproducibility: Validity refers to publication of results that are real, large enough and important enough to matter, and that will generalize to similar settings. We present a Bayesian assessment of a critical component of validity, effect size, taking into account not only the observed data, but also the various sources of bias. The output is a posterior distribution of the true effect size, based on the results of one or more experiments and the biases assumed to exist in those studies. We demonstrate with the simple but real and important example of claimed demonstrations of ESP.

SpeakerShiffrin, Richard
Author 1Shiffrin, Richard
Indiana University
TitleA Bayesian assessment of network similarity
AbstractConsider two networks, possibly of different sizes, for which we know the numbers of nodes and links that are shared and not shared. If the nodes are not labeled this might require first aligning the two networks. I give a formula for the odds that the numbers of observed shared and not shared nodes and links are generated by random sampling from either 1) a single master network (the union of the two observed networks) or 2) two different but similar networks (they have many nodes and links in common). I propose these odds as a measure of network similarity. The calculations being not feasible for very large networks, I propose reducing the two networks in size proportionally, and using the resultant similarity as a stand-in for the large network similarity.

SpeakerSloman, Steven
Author 1Sloman, Steven
Brown University
TitleIgnorance and the Community of Knowledge
AbstractAsking people to explain how something works reveals an illusion of explanatory depth: Typically, people know less about how things work than they think they do. We overestimate our knowledge of common objects. We similarly overestimate our understanding of political policies. How well do you really understand Obamacare? I will argue that the reason we live in this illusion of understanding is that we live in a community of knowledge, guided by shared intentionality. Our communities understand how things work and we fail to distinguish what we know from the knowledge that resides in other people's heads.

SpeakerSloutsky, Vladimir
Author 1Sloutsky, Vladimir
The Ohio State University
TitleWhen Children Outperform Adults: An Adaptive Nature of Developmental Limitations
AbstractChildhood is often construed as a period of developmental limitations: in almost every aspect of human functioning older children and adults outperform younger participants. However, childhood is also the time of unique opportunities in terms of learning new things. In this talk, I present new findings demonstrating how children's limitations in cognitive control, planning, and executive function result in children outperforming adults on attention and memory tasks. I then discuss an adaptive nature of these limitations and argue that they allow optimizing exploration, something that is necessary for successful learning and cognitive development.

SpeakerSperling, George
Author 1Sperling, George
University of California, Irvine
TitleEither visual motion perception or visual attention or combination.
AbstractPlace holder while evaluating the match between possible talks and group interests

SpeakerZacks, Jeffrey M.
Author 1Zacks, Jeffrey M.
Washington University in Saint Louis
Author 2Wahlheim, Christopher N.
University of North Carolina at Greensboro
Author 3Eisenberg, Michelle L.
Washington University in Saint Louis
TitleThe role of episodic memory in event comprehension
AbstractMemory for the recent past guides comprehension of the present and anticipation of the future. To use an event memory on-line, a comprehender must (1) retrieve the appropriate memory at the right time, and (2) accommodate differences between what happened before and what is happening now. Doing so can, in turn (3) shape encoding for subsequent long-term memory. We propose that the mechanism that unifies these three capacities is prediction: Retrieval of previous memories enables predictions during ongoing comprehension, which in turn shapes subsequent encoding. In three experiments, we examined the consequences of noticing and remembering changes for memory for movies of everyday activities. Retrieval of event memories affected eye movements during online viewing and also affected subsequent memory for the new events. This approach gives a new insight into age-related memory impairment: Older adults were less able to use track and use differences between previous and current events to improve memory encoding.

Contact with questions.