Pretrained AI Models: Performativity, Mobility, and Change

TL;DR

This paper examines the societal and ethical impacts of pretrained AI models, highlighting their performative effects, mobility, and the role of users in shaping technological change, with implications for AI governance.

Contribution

It offers a sociological analysis of pretrained models, emphasizing their performative societal influence, mobility, and the importance of user agency in technological evolution.

Findings

Pretrained models can reinforce societal biases and have performative effects.

Users actively reinterpret and modify pretrained models through fine-tuning and transfer.

Understanding sociological aspects can improve AI governance for fairness and transparency.

Abstract

The paradigm of pretrained deep learning models has recently emerged in artificial intelligence practice, allowing deployment in numerous societal settings with limited computational resources, but also embedding biases and enabling unintended negative uses. In this paper, we treat pretrained models as objects of study and discuss the ethical impacts of their sociological position. We discuss how pretrained models are developed and compared under the common task framework, but that this may make self-regulation inadequate. Further how pretrained models may have a performative effect on society that exacerbates biases. We then discuss how pretrained models move through actor networks as a kind of computationally immutable mobile, but that users also act as agents of technological change by reinterpreting them via fine-tuning and transfer. We further discuss how users may use pretrained models in malicious ways, drawing a novel connection between the responsible innovation and user-centered innovation literatures. We close by discussing how this sociological understanding of pretrained models can inform AI governance frameworks for fairness, accountability, and transparency.