We do not directly observe the internal qualities of others so we must infer them from behavior. Although classic attribution theories agree that we consider situational pressures when estimating such internal qualities, one of the best-known results in psychology is that we are prone to a correspondence bias: That we draw inferences from behavior, even when we know that the situation has constrained the action. Dozens of theoretical accounts have sought to explain this result, with the most famous being the proposal that we commit a fundamental attribution error: We are systematically biased to underappreciate the influence of external factors and thus overattribute behavior to disposition. Although there remains disagreement about why we attribute constrained behavior to disposition, most researchers agree that this tendency is in fact an error. We propose that the social judgments made in classic attitude attribution studies have been widely interpreted as reasoning errors only because they have been compared to an inappropriate benchmark, predicated on the assumption of deterministic dispositions and situations. Building from earlier probabilistic accounts, we review classic results that demonstrate that social inferences are consistent with unbiased probabilistic attribution of the influence of situations and dispositions in an uncertain world.

Author(s): Walker, Drew; Vul, Ed | Abstract: How do people assign credit for others’ actions? The Correspondence Bias — a classic bias in social psychology — purports that people are predisposed to attribute behaviors to dispositional, rather than situational, factors. However, recent work suggests that the pattern of data cited as evidence of a bias may be a natural consequence of attribution under uncertainty. Here we devise a novel “Bucket-Toss” task in which we can independently and parametrically manipulate and measure situation and disposition pressures to evaluate whether attribution to dispositions and situations are consistent with probabilistic inference. We find that as the strength of the situation or disposition is varied, attributions to the other (unobserved) cause follow roughly symmetric patterns of graded attribution. Together, these results confirm that social attribution appears to be largely consistent with unbiased inference under uncertainty.

Others’ internal qualities (e.g. dispositions, attitudes) are not directly observable so we must infer them from behavior. Classic attribution theories agree that we consider both internal qualities and situational pressure when making these judgments. However, one of the most well known ideas in psychology is that social judgments are biased, and we tend to underestimate the pressure that situations exert and overestimate the influence of disposition (known as the Fundamental Attribution Error). We propose that the social judgments made in classic studies of attribution have been interpreted as biased only because they have been compared to an inappropriate benchmark of rationality predicated on the assumption of deterministic dispositions and situations. We show that these results are actually consistent with the behavior of a simple ideal Bayesian observer who must reason about uncertain and probabilistic influences of situations and dispositions.

Although many investigations of visual summary representations ("ensemble statistics") have focused on how people compute the central tendency of stimuli such as average set size (e.g. Ariely, 2001), orientation (e.g. Parks, et al., 2001), or facial emotion (e.g. Haberman & Whitney, 2009), less attention has been given to representations of set heterogeneity. People rapidly extract set variance (Michael, et al., 2013), and the variance of a set affects how ensembles are averged (Corbett et al., 2012; Fouriezos et al., 2008; Im & Halberda, 2013). We investigated the ability to detect changes in the variance of circle sizes across sets, using a staircase algorithm. On each trial subjects (n = 23) were presented first with a pedestal display of circles followed by a test display, and had to judge if the variance of the circle sizes (the logarithm of the circle diameter) of the test display was the same as the pedestal set, or if it had changed (the mean was held constant). In one block of 200 trials the changed test variance increased compared to the pedestal variance, while in the other block of 200 trials the changed test variance decreased compared to the pedestal (block order was counterbalanced). We found that people could detect smaller differences between the pedestal and test variance when the variance had decreased, compared to equivalent changes when the variance increased. Meeting abstract presented at VSS 2015.

Previous studies indicate that performance on visual discrimination tasks is enhanced after a delay involving sleep, a result that has been interpreted as reflecting sleep consolidation (Karni, et al., 1994). In the motor domain, however, sleep gains are eliminated when fatigue-reducing spaced practice is used (Rickard, et al., 2008). We applied analogous methodology to test sleep enhancement in a classic texture discrimination task in which subjects indicate whether three slanted lines embedded in an array of lines have a vertical or horizontal orientation. Improvement in this task is measured by a decreased stimulus-to-mask onset asynchrony (SOA), here estimated using a staircase algorithm. Subjects trained on a TDT in the evening and were retested after a 24-hour delay involving normal sleep. Fifteen subjects trained using a standard massed training paradigm involving 192 trials with no breaks. Another 15 trained using an optimized training paradigm. In that condition there were 20s break after every 12 trials to reduce fatigue, and the 192 training trials were randomly interspersed with dummy trials which have been shown to mitigate adaptation effects thought to impair learning (Harris, Gliksberg & Sagi, 2012). Performance in the standard training paradigm was indeed improved on retest. The average SOA (109.6 ms) from the first 8 blocks of Session 2 was significantly decreased compared to the average SOA (137.7 ms) from the last 8 blocks of Session 1, t(1) = 3.45, p = .0014. However, in the optimized training group we found no evidence of enhancement after sleep, t(14) = 0.18, p= .52. The group by session interaction was highly significant, F(1, 28) = 8.4, p = .007. Retinotopic specificity was observed for both groups, indicating that training was sufficient to induce neural changes. These findings invite a theoretical reinterpretation of prior results demonstrating enhanced visual discrimination performance following sleep. Meeting abstract presented at VSS 2014

We give a summary report of an exploratory computational experiment using an “Intuitive Physics Engine” to model developmental discovery processes of mass and momentum as latent physical properties of an object. By this we address parts of the question which knowledge adults and infants must have to perform everyday qualitative physical reasoning. Also, we test whether a computational physics engine can be used as a valid model for this knowledge (and the corresponding acquisition process) in an AI context. We first reproduce earlier findings from qualitative physical reasoning giving evidence of the feasibility of our approach, and then investigate children’s acquisition of an understanding of mass as latent physical property based on observations of object collisions.

Among adults, arithmetic training-transfer studies have documented a high degree of learning specificity. Provided that there is a delay of at least 1 day between training and testing, performance gains do not transfer to untrained problems, nor do they transfer to complement operation-inverted problems (e.g., gains for 4 + 7 = _ do not transfer to the complement subtraction problem, 11 - 4 = _, or vice versa). Here we demonstrate the same degree of learning specificity among 6- to 11-year-old children. These results appear to rule out, for the current training paradigm, operation-level procedural learning as well as any variant of complement problem mediation that would predict transfer. Results are consistent with either or both of two types of learning: (a) item-level procedural learning and (b) a shift to memory-based performance as predicted by the elemental elements model. These results suggest a developmental pattern such that specificity of learning among children is similar to that among adults. Educational implications are noted. (C) 2013 Elsevier Inc. All rights reserved.

In the research reported here, we found evidence of the cheerleader effect-people seem more attractive in a group than in isolation. We propose that this effect arises via an interplay of three cognitive phenomena: (a) The visual system automatically computes ensemble representations of faces presented in a group, (b) individual members of the group are biased toward this ensemble average, and (c) average faces are attractive. Taken together, these phenomena suggest that individual faces will seem more attractive when presented in a group because they will appear more similar to the average group face, which is more attractive than group members' individual faces. We tested this hypothesis in five experiments in which subjects rated the attractiveness of faces presented either alone or in a group with the same gender. Our results were consistent with the cheerleader effect.

We demonstrate that ensemble coding in the visual system works conjointly with other cognitive mechanisms to produce the "the cheerleader effect," the pop-culture notion that individuals are more attractive when they are in a group. We propose that this effect arises because (1) the visual system automatically computes ensemble representations of faces presented in groups (Haberman & Whitney, 2009), (2) averaged faces are perceived as attractive (Langlois & Roggman, 1990), and (3) individual items are drawn to the ensemble average (Brady & Alvarez, 2010). Together, these results suggest that individual faces in a group will be biased toward a group average, and that group average tends to be more attractive than the individual faces, on average. To test this, we found 100 images of groups of three females (experiment 1) and males (experiment 2) on the internet and cropped them to show the three individuals together, or the same three individuals alone. Participants rated the attractiveness of these 300 faces, once presented in a group, and once presented individually. Consistent with the "Cheerleader Effect," female and male faces were rated as more attractive when they were presented with other faces then when they were presented alone (Expt 1: t(24)= 2.53, p <.05; Expt 2: t(18) = 2.13, p <.05 ). We assessed the magnitude of this effect within individuals by assessing how many standard deviations higher a face is rated in a group than alone. In both experiments we found an effect size of about 1/20[sup]th[/sup] of a standard deviation (Expt 1: 5.59%, Expt 2: 5.56%). These findings indicate that automatic averaging of faces produces a summary representation that is more attractive than the faces from which it was derived, and that this representation biases the individual faces in the set to be perceived as more attractive. Meeting abstract presented at VSS 2013

Children are exposed to multiple training tasks that are intended to support acquisition of basic arithmetic skills. Surprisingly, there is a scarcity of experimental research that directly compares the efficacy of those tasks, raising the possibility that children may be spending critical instructional time on tasks that are not effective. We conducted an experiment with 1st through 6th grade children comparing two arithmetic training tasks that are widely used: answer production training and fact triangle training. Results show that answer production training produces substantial fluency gains, whereas fact triangle training does not. Further, we show that, despite theoretical considerations that suggest otherwise, fact triangle training does not produce more flexibly applicable learning. Implications for memory representation, arithmetic fluency training, and broader educational strategy are discussed. (C) 2013 Society for Applied Research in Memory and Cognition. Published by Elsevier Inc. All rights reserved.

It has often been asserted, by both men and women, that men are funnier. We explored two possible explanations for such a view, first testing whether men, when instructed to be as funny as possible, write funnier cartoon captions than do women, and second examining whether there is a tendency to falsely remember funny things as having been produced by men. A total of 32 participants, half from each gender, wrote captions for 20 cartoons. Raters then indicated the humor success of these captions. Raters of both genders found the captions written by males funnier, though this preference was significantly stronger among the male raters. In the second experiment, male and female participants were presented with the funniest and least funny captions from the first experiment, along with the caption author's gender. On a memory test, both females and males disproportionately misattributed the humorous captions to males and the nonhumorous captions to females. Men might think men are funnier because they actually find them so, but though women rated the captions written by males slightly higher, our data suggest that they may regard men as funnier more because they falsely attribute funny things to them.