ETD

• algorithmic bias
• bias
• desgin fiction
• minicode
• toolkit

In today's data-driven landscape, artificial intelligence models play a central role in business strategies and decision-making processes. These systems are widely used in contexts where rapid and complex choices must be made. However, when talking about AI models we must consider these systems as black-boxes. This add a layer of complexity and risk particularly when these models are shaped by biased dataset.
Bias becomes a fundamental issue in evaluating reliability and fairness of AI systems. Many examples have shown that biased datasets can lead systematically outcomes, which, if embedded in decision-making process, may have serious social and ethical consequences.
To address the risks of bad decision-making, it is essential to intervene at the source before using tools such as AI models. Rather than correcting errors produced by automated systems (which can be expensive for companies in terms of time and money) the most effective and efficient approach is to set a reflective mindset easily in the decision-making process. This enables organization and companies to design more ethically sound, context-aware strategies, grounded in a deeper awareness of cognitive and structural bias.

This thesis aims to explore how tools such as MiniCode can be employed to identify, detect, reflect upon and mitigate a group of cognitive and algorithmic bias: metti i nomi during decision-making processes, in organizational and workshop contexts. This work also highlights the importance of integrating tools and strategies from the field of design fiction into decision-making processes, specifically using the MiniCode deck of cards. Design-fiction offers a new approach for this kind of task, providing stakeholders an accessible and efficient way to identify bias and problems in their ideas. By focusing on the imaginative exploration of future scenarios supported by the guided use of cards, the design fiction methodology could bridge the gap between ideation and implementation, helping to develop critical, actionable insights before moving into the creation phase.
To improve the quality and the effectiveness of MiniCode cards and their usability in workshops, I developed the idea of a guiding framework, such as a guide sheet, a visual and conceptual tool designed to guide and support people during the activity. The objective is to provide a structured a guide for users. By the end of the process, participants will have as output of the activity a structured narrative sequence of insights and decision, in other words a card-based path that can inspire further reflection and guide the idea development.
The structure of my work is organized and in 5 chapters:

1. An introduction to cognitive biases, with a particular focus on three specific types: intersectionality bias, anchoring bias, and confounder bias. AlsoAn exploration of the risks associated with neglecting cognitive biases throughout the decision-making process and the creation of AI models which have been built by biased datasets, illustrated through real-world examples.
2. A biref introduction to Design Fiction, its application, its porpuses and limitations.
3. A presentation of the MiniCode tool, including its structured, purpose, and practical application and usage of ethic cards during workshops. A description of Floridi's framework and a disucssion about the limitations of the ethic cards.
4. This chapter introduces a proposed guiding framework throughout a new board designed to support participants throughout the activity, and also a new deck of cards based on bias.
5. The last chapter is dedicated to the proposal of testing MiniCode effectiveness in real-world workshop in order to observe how it can be used to mitigate biases. and also to compare the older version with the new version and see which version is more effective. Additionally , a follow-up for the combination of Minicode toolkit and another powerful tool called Blockly bias is proposed.