Board of Dinner - XAI for Empathy @ DDW

A collaborative work with Deep Food Ltd

Board of Dinner - XAI for Empathy project is an Interactive Menu/Recipe Generator Demonstrating the Potential of AI as an Explainable, Transparent, and Controllable Tool to Foster Empathy.

Can Explainable AI (XAI) gear us toward an empathetic future?

Step into the "Board of Dinner" AI menu generator, an interactive experience where you elect each board member, see a transparent ‘board meeting’ process, get inspired for your next meal, and perhaps discover the inner dialogue of others.

XAI for Empathy project is an invitation to reimagine AI's explainability as a force for empathy. Co-developed by the creative collective Deep Food and Joseph Wan (MA in Data-Driven Design, University of Applied Sciences Utrecht), the project presents an interactive installation called "Board of Dinner". The work features a ‘multi-agents AI’-powered Menu/Recipe Generator designed to be customisable in reflecting various values and transparent in exposing its thinking process.

In a world where AI is often seen as performance-driven yet opaque, the Board of Dinner explores alternatives for unfolding AI's non-human ‘thinking’ mechanism in terms relatable to humans. Stepping into the installation, you will be invited to elect each member of your ‘Board of Dinner’. The ‘Board’ will make a suggestion for your next meal. You can try to include as many perspectives in your board as possible or even the opposite. As the Board generates your personalised menu suggestions alongside their meeting record, you may discover a unique understanding of your choice of human values from the lens of AI.

To cook or not to cook; the experience starts with food and leads to issues beyond—it's about exploring the potential of AI to bridge human limitations in understanding and empathy.

Can we steer AI to enhance trust and mitigate biases through transparency, or should we beware of the other way around? Join our creative experimentation and speculate on the good and bad of XAI together!


Link to all Board of Dinner results, including menu/recipe, alongside meeting minutes of every 'board meeting'.

[https://drive.google.com/drive/folders/1bK6C9leOLLr6C--PxCOWZRrcay4eYEt0?usp=sharing](https://drive.google.com/drive/folders/1bK6C9leOLLr6C--PxCOWZRrcay4eYEt0?usp=sharing)

[Links to all project profile pages and media coverages](https://www.deepfooddesign.com/boardofdinner)

At the core of the Board of Dinner project lies a custom-built multi-agent AI system, where each agent is designed with unique personalities, priorities, and background narratives.

Participants form a "board" by selecting a group of these AI agents, each of whom contributes a dish suggestion based on shared ingredients. The agents then evaluate each other’s proposals using a scoring system, simulating negotiation and preference alignment. After this initial vote, each agent is given a chance to revise the dish to better reflect their values which introduce a second layer of deliberation and personalization. Resulting in a final dish that represents a collective consensus among diverse viewpoints.

The final recipe is converted into a visual prompt and passed to an AI image generator, producing a fictional meal photo. A complete record—documenting the agents' dialogue, votes, and the generated image—is compiled into full meeting minutes, allowing the participant to keep a transparent record of the entire decision-making process. This system is flexible and can be deployed in various contexts, especially for decisions that require input from multiple perspectives.