As generative AI systems increasingly participate in creative processes, their limitations in perspective diversity and explainable reasoning become evident. Most current tools—such as ChatGPT, Notion AI, and Midjourney—assist users in idea generation or refinement, yet they rarely simulate the cognitive plurality inherent to human collaboration. This research explores how multi-role reasoning frameworks can enhance human–AI co-creation by introducing structured diversity and reflective synthesis. Drawing inspiration from De Bono’s Six Thinking Hats, the study develops BrainTeam, an interactive whiteboard prototype that allows users to engage with AI agents representing different cognitive roles—creative, critical, analytical, emotional, and synthetic. Each role contributes distinct reasoning patterns, while a central “Blue Hat” agent integrates them into coherent conclusions. Through a mixed-method approach combining surveys, concept testing, and user evaluations, the study identifies key user needs across creative disciplines: balancing divergent exploration and convergent synthesis, maintaining transparency of reasoning, and fostering explainable collaboration with AI. The final prototype, implemented using React Flow and a Node.js orchestration layer, demonstrates how role-based dialogue can make AI-assisted ideation more structured, participatory, and cognitively transparent. Findings highlight the potential of multi-role frameworks to transform AI from a task executor into a collaborative thinking partner, opening new directions for explainable, human-centered AI design.
Con la crescente partecipazione dei sistemi di intelligenza artificiale generativa ai processi creativi, emergono i loro limiti in termini di diversità di prospettive e di capacità di ragionamento spiegabile. La maggior parte degli strumenti attuali — come ChatGPT, Notion AI e Midjourney — supporta gli utenti nella generazione o nel perfezionamento delle idee, ma raramente simula la pluralità cognitiva tipica della collaborazione umana. Questa ricerca esplora come i framework di ragionamento multi-ruolo possano migliorare la co-creazione uomo–AI introducendo una diversità strutturata e una sintesi riflessiva. Ispirandosi al modello dei “Six Thinking Hats” di De Bono, lo studio sviluppa BrainTeam, un prototipo di lavagna interattiva che consente agli utenti di interagire con agenti AI rappresentanti diversi ruoli cognitivi — creativo, critico, analitico, emotivo e sintetico. Ogni ruolo contribuisce con un proprio schema di ragionamento distinto, mentre un agente centrale “Blue Hat” integra i risultati in conclusioni coerenti. Attraverso un approccio metodologico misto che combina questionari, test di concetto e valutazioni degli utenti, la ricerca identifica i principali bisogni degli utenti nei contesti creativi: bilanciare l’esplorazione divergente e la sintesi convergente, mantenere la trasparenza del ragionamento e promuovere una collaborazione spiegabile con l’AI. Il prototipo finale, implementato utilizzando React Flow e uno strato di orchestrazione Node.js, dimostra come il dialogo basato sui ruoli possa rendere l’ideazione assistita dall’AI più strutturata, partecipativa e cognitivamente trasparente. I risultati evidenziano il potenziale dei framework multi-ruolo nel trasformare l’AI da semplice esecutore di compiti a partner di pensiero collaborativo, aprendo nuove direzioni per un design dell’intelligenza artificiale più spiegabile e centrato sull’uomo.
BrainTeam: a multi-role AI collaboration framework enhancing creativity
Yi, Chunhan
2024/2025
Abstract
As generative AI systems increasingly participate in creative processes, their limitations in perspective diversity and explainable reasoning become evident. Most current tools—such as ChatGPT, Notion AI, and Midjourney—assist users in idea generation or refinement, yet they rarely simulate the cognitive plurality inherent to human collaboration. This research explores how multi-role reasoning frameworks can enhance human–AI co-creation by introducing structured diversity and reflective synthesis. Drawing inspiration from De Bono’s Six Thinking Hats, the study develops BrainTeam, an interactive whiteboard prototype that allows users to engage with AI agents representing different cognitive roles—creative, critical, analytical, emotional, and synthetic. Each role contributes distinct reasoning patterns, while a central “Blue Hat” agent integrates them into coherent conclusions. Through a mixed-method approach combining surveys, concept testing, and user evaluations, the study identifies key user needs across creative disciplines: balancing divergent exploration and convergent synthesis, maintaining transparency of reasoning, and fostering explainable collaboration with AI. The final prototype, implemented using React Flow and a Node.js orchestration layer, demonstrates how role-based dialogue can make AI-assisted ideation more structured, participatory, and cognitively transparent. Findings highlight the potential of multi-role frameworks to transform AI from a task executor into a collaborative thinking partner, opening new directions for explainable, human-centered AI design.| File | Dimensione | Formato | |
|---|---|---|---|
|
thesis.pdf
accessibile in internet per tutti
Descrizione: This thesis explores how multi-role reasoning frameworks can enhance human–AI co-creation. Inspired by De Bono’s Six Thinking Hats, it presents BrainTeam—an interactive whiteboard prototype that enables diverse AI perspectives to support structured, explainable, and collaborative creativity.
Dimensione
123.83 MB
Formato
Adobe PDF
|
123.83 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/247547