Towards a Sociological Conception of AI: Beyond Algorithms and Toward Social Agency

TL;DR

Artificial Intelligence is no longer just a suite of algorithms; it is a "non-human social actor" embedded in our daily lives. This paper argues that sociology has historically ignored AI as a social phenomenon, focusing instead on its technical utility. To bridge this gap, the authors propose a new research agenda to define AI based on societal necessity and human interaction rather than just neural network capabilities.

Background Positioning

In the academic landscape, this work is a critical theoretical intervention. It bridges the gap between Human-Computer Interaction (HCI) and Classical Sociology, advocating for a shift from "Artificial Social Intelligence" (using AI for sociology) to a "Sociology of AI" (studying AI as part of society).

The Problem: The Technology-Society Disconnect

Most current definitions of AI are grounded in Technological Possibilities. We define AI by its architecture—transformers, machine learning, or processing power. However, the authors point out a glaring blind spot:

• Sociological Neglect: Traditional sociology often "theorizes non-human actors out of existence," treating them as mere background environment.
• Micro-focused HCI: While Computer Science studies specific interactions (e.g., how a user talks to a bot), it often lacks a generalizable framework for how these tools change the fabric of society or labor structures.

The researchers highlight a pivotal moment: Mark Zuckerberg’s 2018 Senate testimony. The word "AI" was used significantly more often than "trust" or "transparency," signaling that AI has become a political and social shield, yet we lack the conceptual tools to hold this "actor" accountable.

Methodology: Deciphering the "Social Machine"

The authors propose a robust four-step workflow to build this new conception:

1. Discourse Analysis

By analyzing media and online forums, they aim to map out "common-sense conceptualizations." This identifies the gap between what a developer thinks AI is and what a user fears or expects it to be.

2. Observational Studies: AI in the Wild

The core of the methodology lies in video-based Ethnomethodology.

• Driverless Shuttles: How do groups of humans negotiate space with an autonomous vehicle?
• Chatbots & Games: How is "practical sense-making" achieved when the interlocutor isn't human?

(Note: Figure from the paper illustrating the intersection of technological and social perception)

3. Expert Synthesis

By interviewing both AI developers and Social Scientists, the project aims to create a "Coordinate System" for AI. This matrix would allow us to evaluate AI not just by its FLOPs or Accuracy, but by its social agency and impact on human uniqueness.

Key Insights: Making AI "At Home" in the World

A profound takeaway from the methodology section is the reference to Harvey Sacks: Any novel technological object is "made at home in the world that has whatever organization it already has."

This means AI doesn't create a new world; it inserts itself into existing human social practices. If a chatbot is rude, or an algorithm is biased, it is because it has "made itself at home" in a social structure that already contains those elements.

Critical Analysis & Conclusion

The value of this research lies in its Innovation Potential. Currently, social impact is often an afterthought (Ethics/Safety teams) in the AI development cycle.

Takeaways for the Industry:

• From Post-Hoc to Proactive: Sociology should guide AI development from the start, ensuring systems meet "societal necessities" rather than just "technical milestones."
• Human Uniqueness: As AI performs more human-like tasks, sociology helps us redefine what is "distinctly social" about human behavior.

Limitations: As a prospective research agenda, the paper lacks empirical "hard data" results. Its success depends on whether the AI technical community is willing to adopt a "sociological matrix" in their design process.

Future Outlook: The "Fourth Industrial Revolution" is not just about robots; it's about the transformation of communication. Capturing the "Sociological Conception of AI" is the only way to ensure that as we build more intelligent machines, we don't accidentally dismantle the social structures that make us human.