Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Policy or Community?: Supporting Individual Model Creators' Open Model Development in Model Marketplaces

Eun Jeong Kang, Fengyang Lin, Angel Hsing-Chi Hwang

TL;DR

It is argued that platforms'governance decisions must consider how policy interventions shape the practices and motivations of individual creators, and three regulatory needs shaped by creators' workflows are identified.

Abstract

Lightweight fine-tuning techniques and the rise of 'open' AI model marketplaces have enabled individuals to easily build and release generative models. Yet, this accessibility also raises risks, including the production of harmful and infringing content. While platforms offer policies and responsible AI tools, their effectiveness may be limited, as creators engage with partially open models that vary widely in openness and transparency. To understand how platform governance can better support responsible practices, we conducted semi-structured interviews with 19 individual model creators. We identified three regulatory needs shaped by creators' workflows: reducing downstream harms, recognizing creators' contributions and originality, and securing model ownership. Creators also repurpose RAI tools primarily for self-protection and visibility, and their sense of responsibility is deeply shaped by community norms rather than formal policies. We argue that platforms' governance decisions must consider how policy interventions shape the practices and motivations of individual creators.

Policy or Community?: Supporting Individual Model Creators' Open Model Development in Model Marketplaces

TL;DR

It is argued that platforms'governance decisions must consider how policy interventions shape the practices and motivations of individual creators, and three regulatory needs shaped by creators' workflows are identified.

Abstract

Lightweight fine-tuning techniques and the rise of 'open' AI model marketplaces have enabled individuals to easily build and release generative models. Yet, this accessibility also raises risks, including the production of harmful and infringing content. While platforms offer policies and responsible AI tools, their effectiveness may be limited, as creators engage with partially open models that vary widely in openness and transparency. To understand how platform governance can better support responsible practices, we conducted semi-structured interviews with 19 individual model creators. We identified three regulatory needs shaped by creators' workflows: reducing downstream harms, recognizing creators' contributions and originality, and securing model ownership. Creators also repurpose RAI tools primarily for self-protection and visibility, and their sense of responsibility is deeply shaped by community norms rather than formal policies. We argue that platforms' governance decisions must consider how policy interventions shape the practices and motivations of individual creators.
Paper Structure (56 sections, 3 figures, 2 tables)

This paper contains 56 sections, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Background and Related Work
  3. Open AI Model Development Ecosystem
  4. Challenges to Govern Models on Open AI Model Marketplace Platforms
  5. Responsible AI tools and Communities that Shape Individual Creators' Online Experiences
  6. Responsible AI (RAI) Tools
  7. Community-driven Governance
  8. Methodology
  9. Participants
  10. Recruitment
  11. Demographic Information
  12. Study Design
  13. Interview Protocol Development
  14. Phase 1.
  15. Phase 2.
  16. ...and 41 more sections

Figures (3)

  • Figure 1: Our research problem space showing how individual model creators are situated within the open AI model ecosystem. We focus on individuals who engage with online platforms to build and release models. These creators typically draw on pre-trained open models provided by upstream actors and then fine-tune them using their own approaches ( \ref{['tab:practices']}) to produce models specialized for particular tasks. Their models are subsequently used by downstream users to generate content or to be further remixed into new derivative models.
  • Figure 2: Interview protocols of two phases. Phase 1 included P1-P9, while Phase 2 involved P10-P19.
  • Figure 3: Overview of our findings. Bold line represents current community commitment to support individuals to mitigate compliance challenges