Regardless of expectations for AI systems, interpretability studies seem to be promising towards the future of understanding the mathematical associations of concepts. Breaking down the model to its most atomic representation. This paper explores some of the associations regarding trust. It would be interesting to see if there's a correlation between the embedding of human trust models, the persona vectors of various models, and the ability to jailbreak.
Evaluating LLM Alignment With Human Trust Models
Trust plays a pivotal role in enabling effective cooperation, reducing uncertainty, and guiding decision-making in both human interactions and multi-agent systems. Although it is significant, there is limited understanding of how large language models (LLMs) internally conceptualize and reason about trust. This work presents a white-box analysis of trust representation in EleutherAI/gpt-j-6B, using contrastive prompting to generate embedding vectors within the activation space of the LLM for diadic trust and related interpersonal relationship attributes. We first identified trust-related concepts from five established human trust models. We then determined a threshold for significant conceptual alignment by computing pairwise cosine similarities across 60 general emotional concepts. Then we measured the cosine similarities between the LLM's internal representation of trust and the derived trust-related concepts. Our results show that the internal trust representation of EleutherAI/gpt-j-6B aligns most closely with the Castelfranchi socio-cognitive model, followed by the Marsh Model. These findings indicate that LLMs encode socio-cognitive constructs in their activation space in ways that support meaningful comparative analyses, inform theories of social cognition, and support the design of human-AI collaborative systems.
