@samanthadalal@hci.social

Samantha Dalal @samanthadalal@hci.social

· 23mo ago

Replying to

Want to learn more? Check out our short paper: https://arxiv.org/abs/2411.09102 which will be presented at the NeurIPS EvalEval Workshop on 12/15: We also share an Appendix with pointers for paths forward for participatory AI. [8/8]

arXiv.org

Provocation: Who benefits from "inclusion" in Generative AI?

The demands for accurate and representative generative AI systems means there is an increased demand on participatory evaluation structures. While these participatory structures are paramount to to ensure non-dominant values, knowledge and material culture are also reflected in AI models and the media they generate, we argue that dominant structures of community participation in AI development and evaluation are not explicit enough about the benefits and harms that members of socially marginaliz

1

0

2

0

Samantha Dalal @samanthadalal@hci.social

· 23mo ago

Replying to

GenAI models often create inaccurate, culturally misrepresentative, and insensitive depictions of marginalized communities and what matters to them. This highlights the urgent need for more community input, expertise, and involvement in GenAI development and evaluation. [2/8]

0

6

0

Samantha Dalal @samanthadalal@hci.social

· 23mo ago

Replying to

Many participatory efforts in GenAI rely on "trickle-down" logic: better models will fix stereotypes and preserve material culture. But how, exactly, communities will benefit remains vague. So, how are these engagements and benefits typically structured? [3/8]

0

5

0

Samantha Dalal @samanthadalal@hci.social

· 23mo ago

Replying to

Community members might get paid once for their data, but they’re often cut out of owning or deciding how it’s used. Meanwhile, tech companies profit from the improved models. In the end, communities are sold back “better” tools made possible by their own labor. [4/8]

0

4

0

Samantha Dalal @samanthadalal@hci.social

· 23mo ago

Replying to

Promising communities better GenAI representation ignores that accurate media isn’t just about technical ability. It also depends on social, economic, and political incentives to create and share it. Other users, like marketers or media companies, shape these outcomes too! [5/8]

0

3

0

Samantha Dalal @samanthadalal@hci.social

· 23mo ago

Replying to

Participation in GenAI dev could harm communities. Increasing reliance on AI to make community representations may mean that developers may no longer consult/compensate actual community members themselves, resulting in further marginalization. [6/8]

0

2

0

Samantha Dalal @samanthadalal@hci.social

· 23mo ago

Replying to

Increased visibility in GenAI can bring harms like impersonation. Excluded from tech policymaking, communities must rely on tech companies to protect them from AI harms. But policies that help communities often clash with profit motives. [7/8]

0

1

0