While humans routinely encode and retrieve memories in groups, the bulk of our knowledge of human memory comes from paradigms with individuals in isolation. The primary phenomenon of interest within the relatively new field of collaborative memory is collaborative inhibition: the tendency for collaborative groups to underperform in free recall tasks compared to nominal groups of the same size. This effect has been found in a variety of materials and group compositions (Rajaram & Pereira-Pasarin, 2010). However, the majority of research in this field is guided by verbal theories without formal computational models. In this paper we adapt the Search of Associative Memory (SAM; Raaijmakers & Shiffrin, 1981) model to collaborative free recall. We present a framework to scale SAM to collaborative paradigms with multiple SAM models working together. Our simulation results with the collaborative SAM model suggest that retrieval disruption, responsible for the part-set cuing effect in individuals, is also the cause of collaborative inhibition when multiple models are working together. Our work provides an existence proof that SAM can act as a unified theory to explain both individual and collaborative memory effects, and offers a framework for future predictions of scaling to increased group sizes, shared knowledge, and spread of false memories.

Predictive neural networks, such as word2vec, have seenimpressive recent popularity as an architecture to learndistributional semantics in the fields of machine learning andcognitive science. They are particularly popular because theylearn continuously, making them more space efficient andcognitively plausible than classic models of semantic memory.However, a major weakness of this architecture is catastrophicinterference (CI): The sudden and complete loss of previouslylearned associations when encoding new ones. CI is an issuewith backpropagation; when learning sequential data, the errorsignal dramatically modifies the connection weights betweennodes—causing rapid forgetting of previously learnedinformation. CI is a huge problem for predictive semanticmodels of word meaning, because multiple word sensesinterfere with each other. Here, we evaluate a recentlyproposed solution to CI from neuroscience, elastic weightconsolidation, as well as a Hebbian learning architecture fromthe memory literature that does not produce an error signal.Both solutions are evaluated on an artificial and naturallanguage task in their ability to insulate a previously learnedsense of a word when learning a new one.

Theory of mind is an essential ability for complex social interaction and collaboration. Researchers in cognitive science and psychology have previously sought to integrate theory of mind capabilities into artificial intelligence (AI) agents to improve collaborative abilities (Cuzzolin, Morelli, Cirstea, & Sarahakian, 2020). We introduce the Recurrent Conditional Variational Autoencoder (RCVAE), a novel model which leverages the ability of generative models to learn rich abstracted representations of contextual behaviors to predict behavioral intent from human behavioral trajectories. Advancing on current concept learning models, this model allows for the discovery of latent intent in human behavior trajectories, while maintaining the scalability and performance of generative AI models. We show that in the Overcooked-AI environment, the RCVAE outperforms baseline Long Short-Term Memory (LSTM) models in predicting intent, achieving higher prediction accuracy and greater predictive stability. The implications of these results are significant; the RCVAE's proficiency in learning the relationship between basic actions and resulting contextual behaviors represents a significant advancement in concept learning for behavioral intent prediction.

The primary phenomenon within the field of collaborative memory is collaborative inhibition: the tendency for collaborative groups to underperform in free recall tasks compared to nominal groups of comparable size. Previously, we adapted the Search of Associative Memory (SAM; Raaijmakers & Shiffrin, 1981) model to collaborative free recall and found support for the retrieval disruption hypothesis (Mannering, Rajaram, Shiffrin, & Jones, 2022). In this paper, we investigate another possible cause of collaborative inhibition: memory homogenization. Previous modeling attempts have shown that memory homogenization may contribute to collaborative inhibition (Luhmann & Rajaram, 2015). In this paper, we prevent SAM models from learning during collaborative retrieval which subsequently prevents memory homogenization. We found that even when SAM model memories remained diversified, collaborative inhibition persisted—though the strength of the effect was diminished. These results suggest that both retrieval disruption and memory homogenization may contribute to collaborative inhibition in SAM.