Inferring latent class lexicographic rules from choice data

A lexicographic rule orders multi-attribute alternatives in the same way as a dictionary orders words. Although no utility function can represent lexicographic preference over continuous, real-valued attributes, a constrained linear model suffices for representing such preferences over discrete attributes. We present an algorithm for inferring lexicographic structures from choice data. The primary difficulty in using such data is that it is seldom possible to obtain sufficient information to estimate individual-level preference functions. Instead, one needs to pool the data across latent clusters of individuals. We propose a method that identifies latent clusters of subjects, and estimates a lexicographic rule for each cluster. We describe an application of the method using data collected by a manufacturer of television sets. We compare the predictions of the model with those obtained from a finite-mixture, multinomial-logit model.     

Implicit learning of social predictions

Humans exchange a range of nonverbal social signals in every interaction. It is an open question whether people use these signals, consciously or unconsciously, to guide social behavior. This experiment directly tested whether participants could learn to predict another person’s behavior using nonverbal cues in a single interaction, and whether explicit knowledge of the cue-outcome relationship was necessary for successful prediction. Participants played a computerized game of rock-paper-scissors against an avatar they believed was another participant. Sometimes the avatar generated a predictive facial cue before the play. On these trials, participants’ win-frequency increased over time, even if they did not acquire explicit knowledge of the predictive cue. The degree to which participants could predict the avatar (wins on cued trials) related to their self-reported liking of the avatar. These findings demonstrate the importance of implicit associative learning mechanisms in guiding social behavior on a moment-to-moment basis during interaction.     

Action-effect binding by observational learning

The acquisition of bidirectional action–effect associations plays a central role in the ability to intentionally control actions. Humans learn about actions not only through active experience, but also through observing the actions of others. In Experiment 1, we examined whether action–effect associations can be acquired by observational learning. To this end, participants observed how a model repeatedly pressed two buttons during an observation phase. Each of the buttonpresses led to a specific tone (action effect). In a subsequent test phase, the tones served as target stimuli to which the participants had to respond with buttonpresses. Reaction times were shorter if the stimulus–response mapping in the test phase was compatible with the action–effect association in the observation phase. Experiment 2 excluded the possibility that the impact of perceived action effects on own actions was driven merely by an association of spatial features with the particular tones. Furthermore, we demonstrated that the presence of an agent is necessary to acquire novel action–effect associations through observation. Altogether, the study provides evidence for the claim that bidirectional action–effect associations can be acquired by observational learning. Our findings are discussed in the context of the idea that the acquisition of action–effect associations through observation is an important cognitive mechanism subserving the human ability for social learning.     

Inferring agency from sound

In three experiments we investigated how people determine whether or not they are in control of sounds they hear. The sounds were either triggered by participants’ taps or controlled by a computer. The task was to distinguish between self-control and external control during active tapping, and during passive listening to a playback of the sounds recorded during the active condition. Experiment 1 required detection of a change in control mode within trials. Experiments 2 and 3 introduced a simple rhythm reproduction task that requires discrimination of control modes between trials. The results demonstrate that both sensorimotor cues and perceptual cues are used to infer agency. In addition, there may be further influences of cognitive expectation and/or multimodal integration. In accordance with hierarchical models of intention [e.g., Pacherie, E. (2008). The phenomenology of action: A conceptual framework. Cognition, 107, 179-217] this suggests that the sense of agency is not situated on one specific level of action control but subject to multiple influences.     

A safety mechanism for observational learning

This empirical article presents the first evidence of a “safety mechanism” based on an observational-learning paradigm. It is accepted that during observational learning, a person can use different strategies to learn a motor skill, but it is unknown whether the learner is able to circumvent the encoding of an uncompleted observed skill. In this study, participants were tested in a dyadic protocol in which an observer watched a participant practicing two different motor sequences during a learning phase. During this phase, one of the two motor sequences was interrupted by a stop signal that precluded motor learning. The results of the subsequent retention test revealed that both groups learned the two motor sequences, but only the physical practice group showed worse performance for the interrupted sequence. The observers were consequently able to use a safety strategy to learn both sequences equally. Our findings are discussed in light of the implications of the action observation network for sequence learning and the cognitive mechanisms of error-based observation.     

Perspectives on observational learning in animals

Observational learning is presumed to have occurred when an organism copies an improbable action or action outcome that it has observed and the matching behavior cannot be explained by an alternative mechanism. Psychologists have been particularly interested in the form of observational learning known as imitation and in how to distinguish imitation from other processes. To successfully make this distinction, one must disentangle the degree to which behavioral similarity results from (a) predisposed behavior, (b) increased motivation resulting from the presence of another animal, (c) attention drawn to a place or object, (d) learning about the way the environment works, as distinguished from what we think of as (e) imitation (the copying of the demonstrated behavior). Several of the processes that may be involved in observational learning are reviewed, including social facilitation, stimulus enhancement, several kinds of emulation, and various forms of imitation.     

Demonstrator skill modulates observational aversive learning

Learning to avoid danger by observing others can be relatively safe, because it does not incur the potential costs of individual trial and error. However, information gained through social observation might be less reliable than information gained through individual experiences, underscoring the need to apply observational learning critically. In order for observational learning to be adaptive it should be modulated by the skill of the observed person, the demonstrator. To address this issue, we used a probabilistic two-choice task where participants learned to minimize the number of electric shocks through individual learning and by observing a demonstrator performing the same task. By manipulating the demonstrator’s skill we varied how useful the observable information was; the demonstrator either learned the task quickly or did not learn it at all (random choices). To investigate the modulatory effect in detail, the task was performed under three conditions of available observable information; no observable information, observation of choices only, and observation of both the choices and their consequences. As predicted, our results showed that observable information can improve performance compared to individual learning, both when the demonstrator is skilled and unskilled; observation of consequences improved performance for both groups while observation of choices only improved performance for the group observing the skilled demonstrator. Reinforcement learning modeling showed that demonstrator skill modulated observational learning from the demonstrator’s choices, but not their consequences, by increasing the degree of imitation over time for the group that observed a fast learner. Our results show that humans can adaptively modulate observational learning in response to the usefulness of observable information.     

Outcome-based observational learning in human infants

Previous work shows that infants manifest emulation learning in the use of end-state information. Outcome-based emulation has been interpreted as affordance learning or goal attribution. The present paper explores whether these two learning possibilities might be related. In 3 experiments, 17-month-old infants (N = 180) were presented with action outcomes across a variety of contexts and tasks: They observed either the full demonstration or the model's starting and final postures, plus the initial and end states of the object, or the latter portion of the foregoing display, or the end state of the object alone. The tasks included combinatory, noncombinatory, and body movement acts. Infants reproduced observed outcomes most often by observing the full demonstration. A similar effect was attained by exposure to both posture and configuration changes, but the effect was subject to the combinatory nature of the apparatus. In contrast, performance was less efficient after seeing the object's end state alone, suggesting that infants in the previous conditions did not simply emulate in association with the affordances. These findings support the notion that goal attribution based on sensitivity to bodily cues is reliant on the clarity of the affordances of a task.     

Electrophysiological correlates of observational learning in children

Observational learning is an important mechanism for cognitive and social development. However, the neurophysiological mechanisms underlying observational learning in children are not well understood. In this study, we used a probabilistic reward‐based observational learning paradigm to compare behavioral and electrophysiological markers of individual and observational reinforcement learning in 8‐ to 10‐year‐old children. Specifically, we manipulated the amount of observable information as well as children's similarity in age to the observed person (same‐aged child vs. adult) to examine the effects of similarity in age on the integration of observed information in children. We show that the feedback‐related negativity (FRN) during individual reinforcement learning reflects the valence of outcomes of own actions. Furthermore, we found that the feedback‐related negativity during observational reinforcement learning (oFRN) showed a similar distinction between outcome valences of observed actions. This suggests that the oFRN can serve as a measure of observational learning in middle childhood. Moreover, during observational learning children profited from the additional social information and imitated the choices of their own peers more than those of adults, indicating that children have a tendency to conform more with similar others (e.g. their own peers) compared to dissimilar others (adults). Taken together, our results show that children can benefit from integrating observable information and that oFRN may serve as a measure of observational learning in children.     

Generalization of action knowledge following observational learning

Both observational and physical practices support the acquisition of motor skill knowledge in the form of spatiotemporal coordination patterns. The current experiment examined the extent that observation and physical practice can support the transfer of spatiotemporal knowledge and amplitude knowledge associated with motor skills. Evidence from a multijoint limb task revealed that knowledge about spatiotemporal patterns (relative phase) acquired by observers and models can be generalized exceptionally well within the trained arm (right) and across to the untrained arm (left). Transfer of relative phase occurred even when untrained combinations of joint amplitudes were required. This indicates that observation and physical practice both lead to the development of an effector-independent representation of the spatiotemporal knowledge in this task. Both observers and models showed some transfer of the relative amplitude knowledge, with observers demonstrating superior transfer for both a trained and untrained-arm transfer test, while the models were limited to positive transfer on an untrained-arm transfer test. The representation of movement amplitude knowledge is effector-independent in this task, but the use of that knowledge is constrained by the specific practice context and the linkage between the elbow and wrist.     

Optimistic biases in observational learning of value

Action-outcome contingencies can be learnt either by active trial-and-error, or vicariously, by observing the outcomes of actions performed by others. The extant literature is ambiguous as to which of these modes of learning is more effective, as controlled comparisons of operant and observational learning are rare. Here, we contrasted human operant and observational value learning, assessing implicit and explicit measures of learning from positive and negative reinforcement. Compared to direct operant learning, we show observational learning is associated with an optimistic over-valuation of low-value options, a pattern apparent both in participants’ choice preferences and their explicit post-hoc estimates of value. Learning of higher value options showed no such bias. We suggest that such a bias can be explained as a tendency for optimistic underestimation of the chance of experiencing negative events, an optimism repressed when information is gathered through direct operant learning.     

Inferring psychological significance from physiological signals

A century has passed since the publication of William James's Principles of Psychology, yet most of the questions James raised about the relation between physiological events and molar psychological or behavioral processes, such as emotion, remain unanswered. The sluggish progress in capitalizing on physiological signals to address general psychological questions is due in part to shortcomings in the quantification of physiological signals in humans and, perhaps more important, to the way in which investigators have been thinking about the relation between physiological signals and psychological operations. In this article, we illustrate these points, and we provide a conceptual framework to foster research and analysis of psychological phenomena based on physiological signals. Psychological operations and physiological responses are defined in terms of configural and temporal properties, and psychophysiological relations are conceptualized in terms of their specificity (e.g., one-to-one versus many-to-one) and their generality (e.g., situation or person specific versus cross-situational and pancultural). This model yields four classes of psychophysiological relations: (a) outcomes, (b) concomitants, (c) markers, and (d) invariants. Finally, the model specifies how to determine whether a psychophysiological relation is an outcome, concomitant, marker, or invariant, and it describes important limitations in inferences of psychological significance based on physiological signals when dealing with each.     

Curved saccade trajectories reveal conflicting predictions in associative learning

We report how the trajectories of saccadic eye movements are affected by memory interference acquired during associative learning. Human participants learned to perform saccadic choice responses based on the presentation of arbitrary central cues A, B, AC, BC, AX, BY, X, and Y that were trained to predict the appearance of a peripheral target stimulus at 1 of 3 possible locations, right (R), mid (M), or left (L), in the upper hemifield. We analyzed as measures of associative learning the frequency, latency, and curvature of saccades elicited by the cues and directed at the trained locations in anticipation of the targets. Participants were trained on two concurrent discrimination problems A+R, AC+R, AX+M, X+M and B+L, BC+L, BY+M, Y+M. From a connectionist perspective, cues were predicted to acquire associative links connecting the cues to the trained outcomes in memory. Model simulations based on the learning rule of the Rescorla and Wagner (1972) model revealed that for some cues, the prediction of the correct target location was challenged by the interfering prediction of an incorrect location. We observed that saccades directed at the correct location in anticipation of the target curved away from the location that was predicted by the interfering association. Furthermore, changes in curvature during training corresponded to predicted changes in associative memory. We propose that this curvature was caused by the inhibition of the incorrect prediction, as previously has been suggested with the concept of distractor inhibition (Sheliga, Riggio, & Rizzolatti, 1994; Tipper, Howard, & Houghton, 2000). The paradigm provides a new method to examine memory interference during associative learning.