Toward Digital Twins: Enhancing the Effectiveness of AI-Assisted Communication in Socially Significant Service Domains

RSF project No. 26-18-00438 dated 02.06.2026 (2026-2028).

Project leader: Olessia Koltsova
Main executors: Yadviga Sinyawskaya, Vera Ignatenko, Elena Artemenko
Participants: Polina Kolmogorova, Galina Oreshina, Anton Surkov, Elina Tsigeman-Gorenko, Sofia Shevtsova
Partner client: Data Analysis and Modelling Department, VTB Bank

The Laboratory of Social and Cognitive Informatics (HSE University – St. Petersburg) announces the launch of a major research project supported by the Russian Science Foundation (RSF). The project is being carried out by an interdisciplinary team of researchers from the laboratory, whose expertise covers the entire cycle — from sociological theories to writing code for neural networks.

Annotation
The project focuses on identifying patterns of human interaction with conversational artificial intelligences based on large language models (LLMs) and developing tools to optimize such interaction in order to improve the completeness and comprehensibility of information received by users, as well as enhance the quality of decisions made in socially significant domains: (1) shaping educational and career trajectories, (2) health and health preservation, and (3) consumption of financial services.
The relevance of the study is driven by the transformation of social communication and decision-making practices under the influence of conversational AI systems, which act as new artificial social actors and, prospectively, as "digital twins" of users. Significant risks remain, associated with the uncritical delegation of decisions to algorithms, distortion of problem understanding, and growing dependence on AI.
Within the framework of the project, a sequence of interconnected experiments is being implemented to test hypotheses regarding the factors of effective AI-assisted communication. The study compares three levels of language model adaptation — general-purpose, domain-enriched, and personalized — across a range of objective and subjective performance indicators. Upon completion of the work, a middle-range socio-cognitive theory will be formulated, describing the factors of effective human-AI interaction in information seeking and decision-making in socially significant domains.

Scientific Problem
The scientific problem of the research lies in the need to identify patterns of human interaction with conversational AI assistants and to develop tools for optimizing such interaction, ensuring an increase in the completeness and comprehensibility of the information provided, improving the quality of decisions, and enabling more effective functioning of social institutions. This problem lies at the intersection of sociology, cognitive sciences, psychology, and computer science, and cannot be fully resolved within the framework of a single discipline.

Key Project Objectives

1. To obtain experimentally validated knowledge about the effectiveness of service consumers' interaction with general-purpose generative language agents when solving tasks in the following domains: (1) educational and career pathways; (2) health; (3) consumption of highly complex financial and credit services.
2. To develop a prototype of a domain-specific conversational assistant in one or two of the investigated socially significant domains and to obtain empirically confirmed evidence of its comparative effectiveness relative to general-purpose agents.
3. To develop a prototype of a conversational assistant parametrically tailored to specific socio-psychological types of users (in the financial domain) and to obtain validated data on its comparative effectiveness relative to general-purpose agents.
4. To formulate a middle-range sociological theory that conceptualizes individual-AI interaction in information seeking, information processing, and decision-making in socially significant domains, grounded in a body of sociological and cognitive theories.
5. To develop practical recommendations for integrating the most effective architectural and communicative solutions into AI-assisted communication systems for interaction with clients of a qualified customer.

Expected Results
Over the course of three years, the laboratory will present:

1. Prototypes of "smart" AI assistants adapted to specific domains and user psychotypes.
2. fundamental socio-cognitive theory of human-artificial social agent interaction.
3. Open datasets and validated methodologies for measuring the effectiveness of AI communication.
4. A series of publications in leading international and Russian journals (Q1/Q2 level).
5. Implementation of the results into the customer experience management system of an industrial partner.