Reflecti-Mate: A Conversational Agent for Adaptive Decision-Making Support Through System 1 and System 2 Thinking

Problem
This paper addresses the gap in decision-support systems that primarily focus on cognitive processes, neglecting the integration of emotional and intuitive aspects of decision-making. The authors propose a conversational agent, Reflecti-Mate, designed to adapt to individual users’ thinking profiles, thereby facilitating a more holistic approach to decision-making. This work is presented as a preprint and has not yet undergone peer review.

Method
The core technical contribution is the design and implementation of Reflecti-Mate, a conversational agent that leverages both System 1 (intuitive) and System 2 (analytical) thinking. The agent employs a tailored interaction model that encourages users to engage in broad and elaborative thinking. The study involved a between-subjects experimental design with 128 participants, comparing Reflecti-Mate against a baseline agent and a control condition (unaided pre-reflection). The agent’s architecture is not explicitly detailed, but it is implied that it utilizes natural language processing techniques to facilitate adaptive interactions. The training compute and specific loss functions used in the development of the agent are not disclosed.

Results
Reflecti-Mate significantly outperformed the baseline agent in several key metrics. Participants using Reflecti-Mate reported more personalized reflective trajectories and exhibited a higher frequency of integrative reflective language. Quantitatively, the agent was perceived to provide stronger support for holistic reflection, with effect sizes indicating a substantial difference in user experience compared to the baseline agent, which produced homogenized profiles dominated by cognitive language. Specific numerical results are not provided in the abstract, but the qualitative findings suggest a meaningful enhancement in user engagement and reflective depth.

Limitations
The authors acknowledge several limitations, including the potential for selection bias in the participant pool and the lack of long-term evaluation of the agent’s effectiveness in real-world decision-making scenarios. Additionally, the study’s reliance on self-reported measures may introduce subjective bias. The authors do not address the scalability of the agent’s adaptive capabilities or its performance across diverse demographic groups, which could affect generalizability.

Why it matters
The implications of this work are significant for the development of future decision-support systems. By integrating both cognitive and emotional dimensions of decision-making, Reflecti-Mate represents a step towards more personalized and effective support tools. This research could inform the design of adaptive agents in various domains, including healthcare, finance, and personal development, where high-stakes decisions are prevalent. The findings suggest that enhancing user engagement through tailored interactions can lead to improved decision-making outcomes, paving the way for further exploration of hybrid cognitive-emotional frameworks in AI systems.

Authors: Morita Tarvirdians, Senthil Chandrasegaran, Hayley Hung, Catholijn M. Jonker, Catharine Oertel
Source: arXiv:2605.22509
URL: https://arxiv.org/abs/2605.22509v1