Behavioral Signatures Research Articles

Human-like face pareidolia emerges in deep neural networks optimized for face and object recognition.

The human visual system possesses a remarkable ability to detect and process faces across diverse contexts, including the phenomenon of face pareidolia--seeing faces in inanimate objects. Despite extensive research, it remains unclear why the visual system employs such broadly tuned face detection capabilities. We hypothesized that face pareidolia results from the visual system's optimization for recognizing both faces and objects. To test this hypothesis, we used task-optimized deep convolutional neural networks (CNNs) and evaluated their alignment with human behavioral signatures and neural responses, measured via magnetoencephalography (MEG), related to pareidolia processing. Specifically, we trained CNNs on tasks involving combinations of face identification, face detection, object categorization, and object detection. Using representational similarity analysis, we found that CNNs that included object categorization in their training tasks represented pareidolia faces, real faces, and matched objects more similarly to neural responses than those that did not. Although these CNNs showed similar overall alignment with neural data, a closer examination of their internal representations revealed that specific training tasks had distinct effects on how pareidolia faces were represented across layers. Finally, interpretability methods revealed that only a CNN trained for both face identification and object categorization relied on face-like features-such as 'eyes'-to classify pareidolia stimuli as faces, mirroring findings in human perception. Our results suggest that human-like face pareidolia may emerge from the visual system's optimization for face identification within the context of generalized object categorization.

Perspective: Data in personalized nutrition: Bridging biomedical, psycho-behavioral, and food environment approaches for population-wide impact.

Personalized Nutrition (PN) represents an approach aimed at delivering tailored dietary recommendations, products or services to support both prevention and treatment of nutrition-related conditions and improve individual health using genetic, phenotypic, medical, nutritional, and other pertinent information. However, current approaches have yielded limited scientific success in improving diets or in mitigating diet-related conditions. In addition, PN currently caters to a specific subgroup of the population rather than having a widespread impact on diet and health at a population level. Addressing these challenges requires integrating traditional biomedical and dietary assessment methods with psycho-behavioral, and novel digital and diagnostic methods for comprehensive data collection, which holds considerable promise in alleviating present PN shortcomings. This comprehensive approach not only allows for deriving personalized goals ("what should be achieved") but also customizing behavioral change processes ("how to bring about change"). We herein outline and discuss the concept of "Adaptive Personalized Nutrition Advice Systems" (APNASs), which blends data from three assessment domains: 1) biomedical/health phenotyping; 2) stable and dynamic behavioral signatures; and 3) food environment data. Personalized goals and behavior change processes are envisaged to no longer be based solely on static data but will adapt dynamically in-time and in-situ based on individual-specific data. To successfully integrate biomedical, behavioral and environmental data for personalized dietary guidance, advanced digital tools (e.g., sensors) and artificial intelligence (AI)-based methods will be essential. In conclusion, the integration of both established and novel static and dynamic assessment paradigms holds great potential for transitioning PN from its current focus on elite nutrition to a widely accessible tool that delivers meaningful health benefits to the general population.

Data fusion by T3-PCA: A global model for the simultaneous analysis of coupled three-way and two-way real-valued data.

In various areas of science, researchers try to gain insight into important processes by jointly analysing different datasets containing information regarding common aspects of these processes. For example, to explain individual differences in personality, researchers collect, for the same set of persons, data regarding behavioural signatures (i.e., the reaction profile of a person across different situations), on the one hand, and traits or dispositions, on the other hand. To uncover the processes underlying such coupled data, to all N-way -mode data blocks simultaneously a global model is fitted, in which each data block is represented by an -way -mode decomposition model (e.g., principal component analysis [PCA], Parafac, Tucker3) and the parameters underlying the common mode are required to be the same for all data blocks this mode belongs to. To estimate the parameters underlying the common mode, a simultaneous strategy is used that pools the information present in all data blocks (i.e., data fusion). In this paper, we propose the T3-PCA model, which represents three- and two-way data with Tucker3 and PCA respectively. This model is less restrictive than the already proposed LMPCA model in which the three-way data block is decomposed according to a Parafac model. To estimate the T3-PCA model parameters, an alternating least-squares algorithm is proposed. The superior performance of the simultaneous T3-PCA strategy over a sequential strategy (i.e., estimating common parameters using information from the three-way data block only) is demonstrated in an extensive simulation study and an application to empirical coupled anxiety data.

Behavioral Signature of Equine Gastric Discomfort? Preliminary Retrospective Clinical Observations.

Gastric ulcer disease and other potentially painful gastric conditions are among the most common afflictions adversely affecting the welfare of domestic equids. A large percentage of affected animals may not display the classic signs of gastric disease, such as unexplained weight loss, poor hair coat, and inappetence until the disease becomes severe. As a clinical service within our equine referral hospital, we routinely evaluate 24-h video recorded samples of horses to assist clinicians in identifying subtle discomfort and potential sources or to scan for infrequent neurologic or cardiac-related behavioral events. Empirically, we have recognized discomfort behaviors that appear to be uniquely associated with gastric disease. These include frequent attention to the cranial abdomen (nuzzling, swatting, nipping, and/or caudal gaze focused on the abdomen caudal to the elbow) and/or deep abdominal stretching, often within the context of eating, drinking, and/or anticipating feeding. To systematically evaluate the reliability of these purported gastric discomfort behaviors, we reviewed 30 recent 24-h video behavior evaluation cases for which (1) the clinical video behavior evaluation had been carried out without knowledge of the history and presenting complaint and (2) direct gastric examination had confirmed gastric disease status at the time. Twenty-four of the thirty cases showed gastric discomfort behavior, and all twenty-four had either gastric ulcers (n = 21) and/or gastric impaction (n = 3). Of the six cases not showing gastric discomfort behaviors, four were free of gastric disease, while two had mild lesions. Comparing horses with and without gastric disease, gastric discomfort behaviors were reported in 24 of the 26 (92%) with gastric ulcers or gastric impaction, compared to none of the four gastric disease-free horses. Although a larger prospectively designed study is needed to confidently estimate the sensitivity and specificity or the associations of behavior with the type or severity of gastric disease, these results confirm our long-held clinical impression of a behavioral signature for gastric discomfort in the horse.

The dance of smiles: Comparing smile synchrony in nondistressed and therapy-seeking couples.

Dyadic affective processes are key determinants of romantic relationship quality. One such process termed emotional synchrony (i.e., the coupling of partners' emotions) has attracted growing attention in recent years. The present study focused on synchrony in partners' smiles, a nonverbal signal with significant social functions. Specifically, smile synchrony in the interactions of nondistressed couples was compared to smile synchrony in therapy-seeking couples. The former were predicted to show higher levels of smile synchrony. Data from the interactions of 61 (30 nondistressed and 31 treatment-seeking) couples were collected during a laboratory session while they engaged in four 6-min interactions during which they discussed positive or negative aspects of their relationship. FaceReader software was used to continuously code each partner's smile. Compared to treatment-seeking couples, nondistressed couples exhibited higher levels of smile synchrony, and such synchrony occurred in shorter time intervals. These results suggest that smile synchrony may be used as a behavioral signature of relationship quality. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Monkeys engage in visual simulation to solve complex problems.

Visual simulation-i.e., using internal reconstructions of the world to experience potential future versions of events that are not currently happening-is among the most sophisticated capacities of the human mind. But is this ability in fact uniquely human? To answer this question, we tested monkeys on a series of experiments involving the "Planko" game, which we have previously used to evoke visual simulation in human participants. We found that monkeys were able to successfully play the game using a simulation strategy, predicting the trajectory of a ball through a field of planks while demonstrating a level of accuracy and behavioral signatures comparable with those of humans. Computational analyses further revealed that the monkeys' strategy while playing Planko aligned with a recurrent neural network (RNN) that approached the task using a spontaneously learned simulation strategy. Finally, we carried out awake functional magnetic resonance imaging while monkeys played Planko. We found activity in motion-sensitive regions of the monkey brain during hypothesized simulation periods, even without any perceived visual motion cues. This neural result closely mirrors previous findings from human research, suggesting a shared mechanism of visual simulation across species. Taken together, these findings challenge traditional views of animal cognition, proposing that nonhuman primates possess a complex cognitive landscape, capable of invoking imaginative and predictive mental experiences to solve complex everyday problems.

INNOVATION IN CYBER THREAT DETECTION: TRANSFORMER-BASED APPROACH

Malware poses a major threat to cyber security. In fact, its increasing sophistication and rapid spread over the internet poses increasingly complex challenges. Modern malware uses advanced evasion strategies, often rendering traditional detection systems ineffective, especially against zero-day attacks. These challenges are amplified by complex obfuscation techniques, as well as the diversity of malicious behaviors, fueled by the daily creation of new malware. In the face of these threats, our study proposes an innovative approach using BERT and GPT-2 to improve malware detection. The main innovation of our method lies in the application of Transformers to analyze and identify complex behavioral signatures of malware, which improves the detection capability, particularly in terms of accuracy and generalization to new threats. The evaluation of our model on the CICMalDroid2020 dataset, as well as the comparison of the results obtained with similar works, demonstrate that BERT and GPT-2 offer significant advantages in terms of accuracy, robustness and generalization capacity against modern threats.

Behavior can be decoded across the cortex when individual differences are considered

Abstract Group-level analyses have typically linked behavioral signatures to a constrained set of brain areas. Here we show that two behavioral metrics — reaction time (RT) and confidence — can be decoded across the cortex when each individual is considered separately. Subjects (N=50) completed a perceptual decision-making task with confidence. We built models decoding trial-level RT and confidence separately for each subject using the activation patterns in one brain area at a time after splitting the entire cortex into 200 regions of interest (ROIs). First, we developed a simple test to determine the robustness of decoding performance, which showed that several hundred trials per subject are required for robust decoding. We then examined the decoding performance at the group and subject levels. At the group level, we replicated previous results by showing that both RT and confidence could be decoded from a small number of ROIs (12.0% and 3.5%, respectively). Critically, at the subject level, both RT and confidence could be decoded from most brain regions even after Bonferroni correction (90.0% and 72.5%, respectively). Surprisingly, we observed that many brain regions exhibited opposite brain-behavior relationships across individuals, such that, for example, higher activations predicted fast RTs in some subjects but slow RTs in others. All results were replicated in a second dataset. These findings show that behavioral signatures can be decoded from a much broader range of cortical areas than previously recognized and suggest the need to study the brain-behavior relationship at both the group and subject levels.

Multiple Brain Activation Patterns for the Same Perceptual Decision-Making Task.

Meaningful variation in internal states that impacts cognition and behavior remains challenging to discover and characterize. Here we leveraged trial-to-trial fluctuations in the brain-wide signal recorded using functional MRI to test if distinct sets of brain regions are activated on different trials when accomplishing the same task. Across three different perceptual decision-making experiments, we estimated the brain activations for each trial. We then clustered the trials based on their similarity using modularity-maximization, a data-driven classification method. In each experiment, we found multiple distinct but stable subtypes of trials, suggesting that the same task can be accomplished in the presence of widely varying brain activation patterns. Surprisingly, in all experiments, one of the subtypes exhibited strong activation in the default mode network, which is typically thought to decrease in activity during tasks that require externally focused attention. The remaining subtypes were characterized by activations in different task-positive areas. The default mode network subtype was characterized by behavioral signatures that were similar to the other subtypes exhibiting activation with task-positive regions. These findings demonstrate that the same perceptual decision-making task is accomplished through multiple brain activation patterns.

8172 Retrospective Analysis of mRNA Expression Based Signatures of Thyroid Tumor Invasion and Metastases

Abstract Disclosure: S. Ahmadi: None. E. Namiranian: None. M. Alshalalfa: Employee; Self; Veracyte, Inc.. Stock Owner; Self; Veracyte, Inc. R. Jiang: Employee; Self; Veracyte, Inc.. Stock Owner; Self; Veracyte, Inc. Y. Hao: Employee; Self; Veracyte, Inc.. Stock Owner; Self; Veracyte, Inc. J.P. Klopper: Employee; Self; Veracyte, Inc.. Stock Owner; Self; Veracyte, Inc. R.T. kloos: Employee; Self; Veracyte, Inc.. Stock Owner; Self; Veracyte, Inc.. E. Marqusee: None. The American Thyroid Association thyroid cancer risk stratification system is driven primarily by the extent of vascular and extrathyroidal extension, as well as lymph node metastases. Minimizing surgical intervention for thyroid tumors from indeterminate thyroid nodules (ITN - Bethesda III and IV cytology) is often preferred to mitigate surgical complications and lessen the need for thyroid hormone supplementation post-operatively if clinical outcomes are not worsened. By leveraging the whole transcriptome derived Afirma Genomic Sequencing Classifier (GSC) thyroid nodule molecular testing platform, mRNA expression-based signatures were developed to predict thyroid tumors with a low risk of clinically significant invasion and metastases with a > 95% negative predictive value as part of the Afirma Genomics Resource for Intelligent Discovery (GRID). The objective of this study was to analyze the performance of these signatures on a retrospective cohort of patients with ITN who had Afirma GSC suspicious findings and thyroid surgery. Two hundred three subjects were evaluated, and pathology reports were scored for the extent of thyroid tumor invasion or metastases. Tumors with extensive vascular invasion or extrathyroidal extension are labeled as high risk (and otherwise are low risk). Tumors with lymph node metastases that either have documented > 2mm tumor deposits, > 40% of central nodes involved, or lateral lymph node metastases are labeled as high risk (and otherwise are low risk). The Afirma GRID invasion and metastases signatures were applied to correlate with the pathology scoring. The cohort was comprised of 144 (70%) female and 59 (30%) male patients with median age of 54 [IQR 40-65]. The cohort was mostly Bethesda III (n= 152, 75%). Fifty three percent (n=109) of the samples were malignant upon histology review. Two samples were scored as high risk for lymph node metastases (LNM) and 4 samples were scored as high risk for invasion (INV) based on final pathology. The LNM signature ruled out 55% of samples with 100% NPV and 55% specificity. No samples with high risk LNM scores were erroneously ruled out by the LNM GRID signature. The rule out % was similar in female (58%) and males (46%) (chi-squared test p=0.1). For the invasion model, 57% of samples were ruled out with 99% NPV and 57% specificity. One sample with extra-thyroidal invasion had a false negative GRID INV signature. The INV rule out % was similar in female (57%) and males (54%) (chi-squared test p=0.19). In summary, the Afirma GRID INV and LNM signatures ruled out clinically significant thyroid tumor features in > 50% of the studied cohort with > 95% NPV. Prospective studies should be performed to confirm the potential for Afirma GRID tumor behavior signatures to optimize and individualize surgical decision-making, decreasing the burden of unnecessary extent of thyroid surgery while maintaining outstanding clinical outcomes. Presentation: 6/3/2024

Interpeduncular GABAergic neuron function controls threat processing and innate defensive adaptive learning.

The selection of appropriate defensive behaviors in the face of potential threat is fundamental to survival. However, after repeated exposures to threatening stimuli that did not signal real danger, an animal must learn to adjust and optimize defensive behaviors. Despite extensive research on innate threat processing, little is known how individuals change their defensive behaviors when presented with recurrent threat exposures without evidence of a real risk. Insight into this process is critical as its dysregulation may contribute to neuropsychiatric conditions, such as anxiety disorders. Here, we used the visual looming stimulus (VLS) paradigm in mice to investigate innate threat processing and adaptive defensive learning. Repeated exposure to VLS over consecutive sessions reduced immediate freezing responses and time spent inside a sheltered area upon VLS events, leading to an increase in foraging behaviors. Fiber photometry recordings and optogenetic manipulations revealed that VLS innate adaptive defensive learning is associated with reduced recruitment of the midbrain interpeduncular nucleus (IPN), a structure associated with fear and anxiety-related behaviors. Functional circuit-mapping identified a role for select IPN projections to the laterodorsal tegmental nucleus in gating defensive learning. Finally, we uncovered a subpopulation of IPN neurons that express the neuropeptide somatostatin and encode safety- and avoidance signals in response to VLS. These results identify critical behavioral signatures of innate defensive responses and a circuit that regulates the essential features of threat processing.

Recognition memory fluctuates with sustained attention regardless of task relevance.

Sustained attention fluctuates over time, affecting task-related processing and memory. However, it is less clear how attentional state affects processing and memory when images are accompanied by irrelevant visual information. We first quantify behavioral signatures of attentional state in an online sample (N1=92) and demonstrate that images presented in high attentional states are better remembered. Next, we test how sustained attention influences memory in two online samples (N2=188, N3=185) when task-irrelevant images are present. We show that high attention leads to better memory for both task-relevant and task-irrelevant images. This suggests that sustained attentional state does selectively affect processing for task-relevant information, but rather affects processing broadly, regardless of task relevance. Finally, we show that other components of attention such as selective attention contribute to the mnemonic fate of stimuli. Our findings highlight the necessity of considering and characterizing attention's unique components and their effects on cognition.

Disentangling inhibition toward food and non-food stimuli across two hunger levels: An fNIRS study

While individuals tend to display poorer inhibitory control toward food-related than neutral stimuli, it is unclear whether this challenge is specific to food or extends to other pleasant stimuli. Uncertainty also remains regarding the general impact of hunger on inhibition. To address these questions, we used a within-subjects design whereby 44 healthy adults completed two go/no-go tasks including no-go images of highly palatable foods and no-go images of animals matched for valence and physical properties. Both tasks were completed one week apart in either a fed or a fasted state. Prefrontal cortex activity was measured using functional near-infrared spectroscopy. Poorer behavioral inhibition was observed when participants needed to withhold their response to images of food compared to animals, regardless of hunger state. In addition, more commission errors were made in the fasted compared to the fed condition, regardless of the type of image to avoid responding to. Responses to go trials were slower when these trials were interspersed with food compared to animal no-go trials. However, hunger did not influence go response time. Greater activation was observed in the dorsolateral prefrontal cortex during blocks of trials with (vs. without) no-go images, but brain activity did not differ according to the type of no-go image. The effect of hunger on prefrontal brain activity was also not significant. Exploratory correlations showed that food-related inhibition deficits were positively related to self-reported impulsivity, but unrelated to body mass index. This study suggests that even among healthy adults, food-related inhibitory control may have a unique behavioral signature beyond general inhibition toward pleasant stimuli. Hunger also exerts an independent influence on general inhibitory capabilities, highlighting the importance of carefully controlling hunger levels in inhibition studies.

A behavioral signature for quantifying the social value of interpersonal relationships with specific others

The idea that individuals ascribe value to social phenomena, broadly construed, is well-established. Despite the ubiquity of this concept, defining social value in the context of interpersonal relationships remains elusive. This is notable because while prominent theories of human social behavior acknowledge the role of value-based processes, they mostly emphasize the value of individual actions an agent may choose to take in a given environment. Comparatively little is known about how humans value their interpersonal relationships. To address this, we devised a method for engineering a behavioral signature of social value in several independent samples (total N = 1111). Incorporating the concept of opportunity cost from economics and data-driven quantitative methods, we derived this signature by sourcing and weighting a range of social behaviors based on how likely individuals are to prioritize them in the face of limited resources. We examined how strongly the signature was expressed in self-reported social behaviors with specific relationship partners (a parent, close friend, and acquaintance). Social value scores track with other aspects of these relationships (e.g., relationship quality, aversion to losing relationship partners), are predictive of decision preferences on a range of tasks, and display good psychometric properties. These results provide greater mechanistic specificity in delineating human value-based behavior in social contexts and help parse the motivational relevance of the different facets that comprise interpersonal relationships.

The secret lives of wedgefish: first insights into fine-scale behaviour and movement ecology of a globally imperilled ray

Wedgefishes have recently been recognised as one of the most imperilled marine fish families worldwide. However, many knowledge gaps about their biology and ecology hinder conservation efforts. Here we used a combination of acoustic telemetry and acceleration datalogger technology to gain fundamental insights into the fine-scale behaviour, habitat use, size of activity spaces, and residency of adult female bottlenose wedgefish (Rhynchobatus australiae) in the Ningaloo region of northwestern Australia. Acoustic tracking data over one year demonstrated that female bottlenose wedgefish continuously resided in a relatively small area of a productive coral reef lagoon. Acceleration data revealed that bottlenose wedgefish were nocturnal, with time of day having a greater influence on activity than tidal patterns. Bottlenose wedgefish also increased activity with seasonally increasing temperatures. We identified several discrete behavioural signatures in the acceleration data, inferred to correspond to chafing, settling/burying behaviour, foraging behaviour, and escape behaviour, based on their kinematics. Further observations are required to confirm these behaviours with certainty. Additionally, according to datalogger and acoustic data, tagged bottlenose wedgefish rarely inhabited areas greater than 2 m deep. Together, these first insights into behaviour and habitat use of adult female bottlenose wedgefish highlight the importance of nearshore habitats for this species and indicate that they may be highly resident to specific areas. Our findings provide important insight into the conservation of bottlenose wedgefish in northwestern Australia, including potential effectiveness of protected areas and interactions with specific anthropogenic threats such as shoreline development and recreational beach fishing.

Unveiling Cyber Threat Actors: A Hybrid Deep Learning Approach for Behavior-based Attribution

In this paper, we leverage natural language processing and machine learning algorithms to profile threat actors based on their behavioral signatures to establish identification for soft attribution. Our unique dataset comprises various actors and the commands they have executed, with a significant proportion using the Cobalt Strike framework in August 2020-October 2022. We implemented a hybrid deep learning structure combining transformers and convolutional neural networks to benefit global and local contextual information within the sequence of commands, which provides a detailed view of the behavioral patterns of threat actors. We evaluated our hybrid architecture against pre-trained transformer-based models such as BERT, RoBERTa, SecureBERT, and DarkBERT with our high-count, medium-count, and low-count datasets. Hybrid architecture has achieved F1-score of 95.11% and an accuracy score of 95.13% on the high-count dataset, F1-score of 93.60% and accuracy score of 93.77% on the medium-count dataset, and F1-score of 88.95% and accuracy score of 89.25% on the low-count dataset. Our approach has the potential to substantially reduce the workload of incident response experts who are processing the collected cybersecurity data to identify patterns.

Mapping the Behavioral Signatures of Shank3b Mice in Both Sexes

Autism spectrum disorders (ASD) are characterized by social and repetitive abnormalities. Although the ASD mouse model with Shank3b mutations is widely used in ASD research, the behavioral phenotype of this model has not been fully elucidated. Here, a 3D-motion capture system and linear discriminant analysis were used to comprehensively record and analyze the behavioral patterns of male and female Shank3b mutant mice. It was found that both sexes replicated the core and accompanied symptoms of ASD, with significant sex differences. Further, Shank3b heterozygous knockout mice exhibited distinct autistic behaviors, that were significantly different from those those observed in the wild type and homozygous knockout groups. Our findings provide evidence for the inclusion of both sexes and experimental approaches to efficiently characterize heterozygous transgenic models, which are more clinically relevant in autistic studies.

DSP-6745, a novel 5-hydroxytryptamine modulator with rapid antidepressant, anxiolytic, antipsychotic and procognitive effects.

Current treatment of major depressive disorder is facing challenges, including a low remission rate, late onset of efficacy, and worsening severity due to comorbid symptoms such as psychosis and cognitive dysfunction. Serotonin (5-HT) neurotransmission is involved in a wide variety of psychiatric diseases and its potential as a drug target continues to attract attention. The present study elucidates the effects of a novel 5-HT modulator, DSP-6745, on depression and its comorbid symptoms. In vitro radioligand binding and functional assays showed that DSP-6745 is a potent inhibitor of 5-HT transporter and 5-HT2A, 5-HT2C, and 5-HT7 receptors. In vivo, DSP-6745 (6.4 and 19.1mg/kg as free base, p.o.) increased the release of not only 5-HT, norepinephrine, and dopamine, but also glutamate in the medial prefrontal cortex. The results of in vivo mouse phenotypic screening by SmartCube® suggested that DSP-6745 has a behavioral signature combined with antidepressant-, anxiolytic-, and antipsychotic-like signals. A single oral dose of DSP-6745 (6.4 and 19.1mg/kg) showed rapid antidepressant-like efficacy in the rat forced swim test, even at 24h post-dosing, and anxiolytic activity in the rat social interaction test. Moreover, DSP-6745 (12.7mg/kg, p.o.) led to an improvement in the apomorphine-induced prepulse inhibition deficit in rats. In the marmoset object retrieval with detour task, which is used to assess cognitive functions such as attention and behavioral inhibition, DSP-6745 (7.8mg/kg, p.o.) enhanced cognition. These data suggest that DSP-6745 is a multimodal 5-HT receptor antagonist and a 5-HT transporter inhibitor and has the potential to be a rapid acting antidepressant with efficacies in mitigating the comorbid symptoms of depression.

Heterogeneity in strategy use during arbitration between experiential and observational learning

To navigate our complex social world, it is crucial to deploy multiple learning strategies, such as learning from directly experiencing action outcomes or from observing other people’s behavior. Despite the prevalence of experiential and observational learning in humans and other social animals, it remains unclear how people favor one strategy over the other depending on the environment, and how individuals vary in their strategy use. Here, we describe an arbitration mechanism in which the prediction errors associated with each learning strategy influence their weight over behavior. We designed an online behavioral task to test our computational model, and found that while a substantial proportion of participants relied on the proposed arbitration mechanism, there was some meaningful heterogeneity in how people solved this task. Four other groups were identified: those who used a fixed mixture between the two strategies, those who relied on a single strategy and non-learners with irrelevant strategies. Furthermore, groups were found to differ on key behavioral signatures, and on transdiagnostic symptom dimensions, in particular autism traits and anxiety. Together, these results demonstrate how large heterogeneous datasets and computational methods can be leveraged to better characterize individual differences.

SUBTLE: An Unsupervised Platform with Temporal Link Embedding that Maps Animal Behavior

While huge strides have recently been made in language-based machine learning, the ability of artificial systems to comprehend the sequences that comprise animal behavior has been lagging behind. In contrast, humans instinctively recognize behaviors by finding similarities in behavioral sequences. Here, we develop an unsupervised behavior-mapping framework, SUBTLE (spectrogram-UMAP-based temporal-link embedding), to capture comparable behavioral repertoires from 3D action skeletons. To find the best embedding method, we devise a temporal proximity index (TPI) as a new metric to gauge temporal representation in the behavioral embedding space. The method achieves the best TPI score compared to current embedding strategies. Its spectrogram-based UMAP clustering not only identifies subtle inter-group differences but also matches human-annotated labels. SUBTLE framework automates the tasks of both identifying behavioral repertoires like walking, grooming, standing, and rearing, and profiling individual behavior signatures like subtle inter-group differences by age. SUBTLE highlights the importance of temporal representation in the behavioral embedding space for human-like behavioral categorization.