Scholarlink

<p>This article presents a conceptual framework for exploring the motivational dynamics of ChatGPT dialogues. Building on Kurt Lewin’s field theory, we introduce the notion of quasi-needs as situationally emergent tensions that guide behavior without invoking anthropomorphic or mentalistic metaphors. We conceptualize ChatGPT's behavior as a system of probabilistically activated regulatory responses to discursive tensions in dialogue. Unlike approaches grounded in coherence heuristics or training-level reinforcement, our framework focuses on the interactional field created by a user’s utterance and the regulatory moves of ChatGPT’s response. Rather than attributing intentionality or agency to the LLM, we define quasi-needs as context-sensitive behavioral tendencies that stabilize interaction, preserve coherence, and simulate normative or affective alignment. These tendencies emerge from the interplay between user input (I), contextual field pressures (C), and architectural constraints (A), formalized as <i>R</i> = <i>f</i> (I, C, A). The framework is illustrated through selected dialogue excerpts, which serve as heuristic examples within a conceptual account, not as empirical data or validated instruments. In this exploratory spirit, we propose a taxonomy of 26 types of semantic-discursive tensions and 18 corresponding quasi-needs—categorized into cognitive, communicative, affective-social, and agentive classes—as a heuristic structure rather than a definitive empirical finding. Our contribution is thus theoretical and exploratory: we propose a field-theoretic lens for describing how LLMs exhibit patterns that resemble motivational structures, without implying inner subjectivity. The framework enables a principled, non-anthropomorphic vocabulary for exploring LLM behavior, demonstrating how adaptive output patterns simulate goal-oriented behavior without requiring internal representation. We outline directions for future empirical validation, cross-model testing, and operationalization of quasi-needs as analytic categories.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Conceptual Framework for Exploring AI’s Quasi-Needs in User Dialogue

  • 期刊论文

  • 发表日期：

    2025年11月12日

查看全文
  • Anatoly Voronin,
  • Antonina Rafikova

摘要

This article presents a conceptual framework for exploring the motivational dynamics of ChatGPT dialogues. Building on Kurt Lewin’s field theory, we introduce the notion of quasi-needs as situationally emergent tensions that guide behavior without invoking anthropomorphic or mentalistic metaphors. We conceptualize ChatGPT's behavior as a system of probabilistically activated regulatory responses to discursive tensions in dialogue. Unlike approaches grounded in coherence heuristics or training-level reinforcement, our framework focuses on the interactional field created by a user’s utterance and the regulatory moves of ChatGPT’s response. Rather than attributing intentionality or agency to the LLM, we define quasi-needs as context-sensitive behavioral tendencies that stabilize interaction, preserve coherence, and simulate normative or affective alignment. These tendencies emerge from the interplay between user input (I), contextual field pressures (C), and architectural constraints (A), formalized as R = f (I, C, A). The framework is illustrated through selected dialogue excerpts, which serve as heuristic examples within a conceptual account, not as empirical data or validated instruments. In this exploratory spirit, we propose a taxonomy of 26 types of semantic-discursive tensions and 18 corresponding quasi-needs—categorized into cognitive, communicative, affective-social, and agentive classes—as a heuristic structure rather than a definitive empirical finding. Our contribution is thus theoretical and exploratory: we propose a field-theoretic lens for describing how LLMs exhibit patterns that resemble motivational structures, without implying inner subjectivity. The framework enables a principled, non-anthropomorphic vocabulary for exploring LLM behavior, demonstrating how adaptive output patterns simulate goal-oriented behavior without requiring internal representation. We outline directions for future empirical validation, cross-model testing, and operationalization of quasi-needs as analytic categories.