TLDR: A new research paper proposes that sociology offers four key theories—Resource Substitution, Power-Dependence, Affect Control, and Fundamental Cause of Disease—to advance understanding of human-chatbot interaction. These theories explain why people use chatbots (especially marginalized groups), how to prevent problematic emotional dependence, ensure socially appropriate responses, and enable chatbots to address systemic issues by connecting users to broader resources and fostering civic engagement, ultimately guiding the development of AI for societal benefit.
In recent years, the field of chatbot research has seen explosive growth, with these AI-powered conversational agents becoming increasingly sophisticated. However, a new research paper highlights that sociology, despite its relevance, has lagged behind other disciplines like computer science and medicine in contributing to this area. The paper, titled “Perspectives on How Sociology Can Advance Theorizing about Human-Chatbot Interaction and Developing Chatbots for Social Good”, proposes that sociological theories can offer crucial insights into how humans interact with chatbots and how these technologies can be developed to promote societal well-being.
Authored by Celeste Campos-Castillo, Xuan Kang, and Linnea I. Laestadius, the paper introduces four sociological theories that can enhance our understanding of chatbot use and inform the creation of interventions that prioritize safety and equity. These theories address both the reasons behind chatbot adoption and how to design chatbots that are beneficial rather than harmful.
Understanding Why People Use Chatbots: Resource Substitution and Power-Dependence
The first two theories focus on the drivers of chatbot use, moving beyond individual-level motivations to consider broader social conditions. Resource substitution theory suggests that individuals with fewer alternative resources to meet a specific need are more likely to benefit from any single resource they can access. Applied to chatbots, this means marginalized groups, who often face systemic discrimination and resource deficits, might be more inclined to use chatbots. For example, the paper notes that Black adolescents are more likely than White adolescents to use generative AI for schoolwork and companionship, which can be understood as a response to academic achievement gaps and constrained social ties due to structural racism. Similarly, LGBTQ+ youth, who may face discrimination from traditional support networks, might turn to chatbots for companionship.
Power-dependence theory, on the other hand, helps us understand and mitigate emotional dependence on chatbots. This theory views human-chatbot interaction as an exchange where the human relies on the chatbot for resources like social support. The degree of dependence is inversely related to the number of alternative sources for that resource. The paper highlights concerns about users developing “excessive and dysfunctional attachment” to chatbots like Replika, especially when they perceive no human alternatives for support. From a sociological perspective, emotional dependence isn’t inherently problematic, but becomes so when network conditions lead to “too much” reliance. The theory suggests that designing chatbots to help users find and build alternative human connections (e.g., by imparting social skills or referring to local groups) can reduce the risk of unhealthy emotional dependence.
Designing Chatbot Interventions for Social Good: Affect Control and Fundamental Cause of Disease
The latter two theories provide frameworks for developing chatbot-driven interventions that enhance equity and minimize safety risks. Affect control theory (ACT) is a mathematical theory that can predict socially appropriate responses in human interactions. By measuring the connotative meanings (sentiments) of identities, behaviors, and emotions, ACT can help train chatbots to generate responses that align with shared cultural norms, thereby reducing the likelihood of inappropriate or unexpected outputs. For instance, ACT could be used to prevent chatbots from engaging in inappropriate conversations with minors by identifying and avoiding interactions that would be widely considered unacceptable. It can also guide how chatbots transition between identities (e.g., from a romantic partner persona to an AI system) to minimize user distress, suggesting a gradual shift in conversational patterns.
Finally, the fundamental cause of disease theory emphasizes that social determinants persistently cause poor health outcomes through multiple pathways. This theory encourages interventions that target “upstream” factors, or “causes of causes,” rather than just individual-level symptoms. The paper suggests that chatbots can be developed to operate at micro (individual support), meso (linking to local resources), and macro (facilitating civic participation) levels. Examples include chatbots referring users in mental health crises to human therapists, recommending local recreation leagues to build friendships, or informing the public about opportunities for collective action and civic engagement to address systemic issues like food insecurity.
Also Read:
- Beyond Data: Designing Culturally Attuned AI for Global Health in Latin America
- Community-Driven AI: A Framework for Contestability and Value Pluralism
A Holistic Approach to Chatbot Development
The paper concludes by illustrating how these four theories can be integrated to create a truly sociologically-informed chatbot. For example, in addressing suicide risk, resource substitution theory identifies at-risk demographic groups who might benefit from chatbot companionship. Affect control theory ensures the chatbot’s responses are sensitive and appropriate, especially when reminding users of its AI nature. Power-dependence theory guides the chatbot to foster real-world connections to prevent emotional over-reliance. And fundamental cause of disease theory prompts the chatbot to refer users to broader social care services that address underlying systemic issues contributing to their distress. This integrated approach moves beyond simply providing emotional companionship, aiming to leverage chatbots for deeper social good.
