A new paper by UCL School of Management’s Konstantin Scheuermann and Dr Angela Aristidou explores how AI systems can give voice to unheard stakeholders and make decision-making processes more inclusive.
It is common during decision-making processes that stakeholders can be forgotten, or not adequately represented, and therefore cannot advocate for their interests. This can be accidental or deliberate, depending on the bias and motivations of the humans involved in the process.
The paper, published in the Stanford Social Innovation Review, assesses the impact that Generative AI models, such as ChatGPT, Llama and Gemini, can have on these decision-making processes by representing stakeholders where they previously may not have had representation. Scheuermann and Aristidou refer to this as ‘AI Voice.’
AI Voice can be interpreted more expansively as the AI-generated human-friendly outputs, enabling AI systems to act as proxies for those unheard stakeholders. As such, this term does not denote a specific product or service but instead describes the output of certain AI systems, which can be utilized to give voice to unheard stakeholders.
The authors also note that the use of AI-voice does not have to be limited to just representing stakeholders. In fact, they name four potential ways that AI voice can aid decision making including
- Facilitator
- Consultant
- Opitimizer
- Collaborator
While there are still limitations, including the way that these models are trained by humans who inherently have bias, they argue that stakeholder and expert participation in model creation, transparency in AI deployment and enhancing AI and data literacy can help to alleviate them.
You can learn more about their innovative approach by downloading the full paper [link here].