Psychology

Chomsky (1967): "A Review of B. F. Skinner's Verbal Behavior" link

Bayesian Cognition

*Daw and Courville (2007): "The pigeon as particle filter" pdf
summary
This paper proposes that discontinuities in human learning with aggregate Bayesian learning can be explained by a particle filter model of belief formation. Individuals are sure about what they think at any point in time, but what they think changes according to a sequential Monte Carlo sampler. The paper also addresses that problem that resampling or reweighting needed to obtain good aggregate posterior estimates in particle filtering is not possible at the individual level. The authors propose that individuals could either have a correct model of a jumpy world that leads to periodic resampling or an incorrect model of a stable world that gives the same behavior. (Note: this seems a bit hacky.) The authors show that this model can account for some qualitative properties of existing individual and aggregate level data, including slow learning even at the aggregate level, which they claim may be partially due to aggregation without importance weighting. Interestingly, the model with jumps even displays backward blocking at the aggregate level---training a reward on A and B then presenting B makes A unlikely because the individuals who thought it was A jump to B at the presentation of B.

Gershman (2015): "A Unifying Probabilistic View of Associative Learning" link
summary
This paper present a model of associate learning that combines of elements Bayesian learning (in the form of a Kalman filter, where the hidden parameters are the weights that combine with stimuli to form the reward) and reinforcement learning (in the form of temporal difference learning to account for future reward, not just immediate reward). This synthesis helps account for failings of the Rescorla-Wagner model such as (1) the effects of inter-trial stimuli that are presented without reward signals or reinforcement through pairing of stimuli rather than pairing of stimuli and reward (model accomplishes via long-term reward considerations); (2) decreased reinforcement in absence of any stimuli (via increase in uncertainty in a random walking Kalman filter without evidence); and (3) various blocking phenomena (via inference of negative correlations between stimuli weights). The combined model as has unique benefits of explaining blocking phenomena on pairs of stimuli presented without reward. (PK: it's a bit unusual to me to see Bayesian learning being used in this way. While I see the mathematical relationship between the Rescorla-Wagner update and the Kalman filter update, normally I would think of the Kalman filter as tracking reward weights that change over time, while R-W as estimating fixed weights. Which is it?)

Lewandowsky et al. (2009): "The Wisdom of Individuals: Exploring People's Knowledge about Everyday Events using Iterated Learning" pdf
summary
This paper provides further evidence for Bayesian inference in cognition using a technique called iterated learning. They replicate an extensive study of Bayesian inference of Griffiths and Tenenbaum (2006) and show that indeed the Bayesian model fits better than a proposed alternative heuristic model. The prediction task given to experimental participants is of the following form: Given you have been waiting in line for 10 minutes, how long do you think you will continue to wait? In order to elicit priors, they use the participant's answer to this question as the parameter of the next question (the likelihood they use to generate the next sample is uniform between 0 and the previous response). The experimental design used to compare to the alternative model is quite clever.

Sanborn and Griffiths (2007): "Markov Chain Monte Carlo with People" pdf
summary
This paper provides a way to implement MCMC using people as acceptor functions. The authors make use of the relationship between the functional form of the Barker acceptance rule for the Metropolis algorithm and the funcitonal form of the posterior probability of choosing a data point to be from a particular category when two data points are presented: one from the category and one from a uniform distribution. The one from the category plays the role of the previous state of the MCMC chain while the one from the uniformt distribution plays the role of the next proposed value.

Zednik and Jakel (2014): "How does Bayesian reverse-engineering work?" pdf
summary
This paper explores the role of Bayesian models in reverse-engeering human cognition in Marr's sense. The paper argues against Bayesian realism, the view that a Bayesian model fitting human data implies people are actually performing that Bayesian inference in that model. The paper cites evidence showing that alternative models can produce results that look like Bayesian inference. The paper also argues against Bayesian instrumentalism, the view that Bayesian models only provide explanations at the computational level of analysis and not necessarily at the level of algorithmic analysis. The authors' argument against Bayesian instrumentalism is fairly weak. Their argument is basically that Bayesian intrumentalism isn't a very useful perspective for reverse-engineering since under this view, Bayesian models don't inform lower levels of understanding. The authors advocate for pragmatic Bayesianism, which supposes that Bayesian models provide constraints on lower levels of understanding and thus are useful for hypothesis generation.

Cognitive Psychology

Jones et al. (2013): "Report'from'the'NSF'Workshop'on Integrating'Approaches'to'Computational Cognition" pdf
*Kimbrough et al. (2013): "The Evolution of 'Theory of Mind': Theory and Experiments" link
*McCoy et al. (2012): "Why blame Bob? Probabilistic generative models, counterfactual reasoning, and blame attribution" pdf
Pacer et al. (2013): "Evaluating computational models of explanation using human judgments" pdf
*Seyfarth and Cheney (2013): "Affiliation, empathy, and the origins of theory of mind" link

Conversational Dynamics

Choudhury and Basu (2004): "Modeling Conversational Dynamics as a Mixed-Memory Markov Process" pdf
Eagle et al. (2003): "Common Sense Conversations: Understanding Casual Conversation using a Common Sense Database" pdf

Creativity

Amabile et al. (1996): "Assessing the Work Environment for Creativity" link
Zhu et al. (2009): "Human Rademacher Complexity" pdf

Heuristics and Biases

Costello and Watts (2013): "Surprisingly Rational: Evidence that people follow probability theory when judging probabilities, and that biases in judgment are due to noise" pdf
Goldstein and Rothschild (2014): "Lay understanding of probability distributions" pdf
summary
This paper suggests that results reporting people do not have very good intuitive understandings of probability distributions may be exaggerated by experimental design. The authors use a method of having participants draw histograms rather than report predictions and statistics directly. They find that the aggregate performance is very good, which indicates low individual bias, find that this "graphical method" results in much higher performance.

Kahneman (2002): "Maps of Bounded Rationality: Psychology for Behavioral Economics" pdf

Intuitive Theories

Knobe (2010): "Person as Scientist, Person as Moralist" pdf

Perception

Sam Gershman and Yael Niv (2013): "Perceptual Estimation Obeys Occam's Razor" link

Social Cognition

Arieli and Mueller-Frank (2013): "Inferring Beliefs from Actions" link
Baker and Tenenbaum (2014): "Modeling Human Plan Recognition using Bayesian Theory of Mind" pdf
Baker et al. (2008): "Theory-based Social Goal Inference" pdf
Frank (20??): "Learning words through probabilistic inferences about speakers' communicative intentions" pdf
Frank and Goodman (20??): "Inferring word meanings by assuming that speakers are informative" pdf
Goodman et al. (2009): "Cause and Intent: Social Reasoning in Causal Learning" pdf
Goodman and Stuhlmuller (2013): "Knowledge and Implicature: Modeling Language Understanding as Social Cognition" pdf
*Kinzler et al. (2007): "The native language of social cognition" link
Onishi and Baillargeon (2005): "Do 15-Month-Old Infants Understand False Beliefs?" link
Pantelis et al. (2014): "Inferring the intentional states of autonomous virtual agents" link
Shafto et al. (20??): "Learning from others: The consequences of psychological reasoning for human learning" pdf
Smith et al. (2013): "Learning and using language via recursive pragmatic reasoning about other agents" pdf
Ullman et al. (2010): "Help or Hinder: Bayesian Models of Social Goal Inference" pdf Wimmer and Perner (1983): "Beliefs about beliefs" pdf

Social Psychology

*Abelson et al. (1998): "Perceptions of the collective other" pdf
Dasgupta et al. (1999): "Group entitativity and group perception: Associations between physical features and psychological judgment" pdf
Hamilton and Sherman (1996): "Perceiving persons and groups" link
Hamilton et al. (2011): "A Group By Any Other Name—The Role of Entitativity in Group Perception" link
McConnell et al. (1997): "Target entitativity: Implications for information processing about individual and group targets" link
Meltzoff and Moore (1983): "Newborn Infants Imitate Adult Facial Gestures" link
Sani et al. (2007): "Perceived collective continuity: seeing groups as entities that move through time" link
Sherif et al. (1961): "Intergroup Conflict and Cooperation: The Robbers Cave Experiment" link
Tajfel et al. (1971): "Social categorization and intergroup behaviour" link
Peter M Krafft Last modified: Mon Dec 29 14:01:47 EST 2014