The Horizon Europe funded TITAN project headed to Rome, Italy, this month as project partners gathered to reflect on their progress, and work through crucial decisions ahead of the Summer period. The meeting held on the 10th and 11th of May was hosted by Engineering, at the Centro Congress building in the heart of the city.
During the meeting, partners discussed their progress in working towards the overall aim of TITAN, which is to use AI to help raise citizens critical thinking skills in determining disinformation online. From hosting multiple citizen co-creation workshops which help inform the socio-technical framework that underpins the TITAN solution, to a focus on expert stakeholder engagement, partners have created a firm human-centered footing on which to start technical design and development. A foundation that will be continued through the establishment of a Living Lab phase for agile testing sprints in Belgium, Denmark, Italy and Bulgaria.
What is TITAN about?
The TITAN solution, in simple terms, is an intelligent conversational agent (chatbot) that will provide personal coaching to users wanting to question the validity of online content. It achieves it aim through the analysis and annotation of structured dialogues using an open large language model (LLM). LLM's are a type of artificial intelligence (AI) that use deep learning techniques and massive data sets to understand, summarise, generate and predict new content. In essence it can have a 'smart' conversation with a user.
In TITAN the aim is to create a chat agent that asks the user questions using Socratic thinking principles, analyses the users answer according to their level of critical thinking, and poses an appropriate question to help prompt further thinking of the issue at hand.
Exploring results from citizen co-creation workshops in 8 countries helped partners to understand user needs and potential blockers which need to be overcome. Some citizens seemed a little wary of AI solutions where data was concerned. For example, 57% of participants said they preferred not to be obligated to share personal data, a finding that helped partners explore the concept of what constitutes personal data in the context of TITAN, and how to measure trustworthiness of the system.
Other considerations, explored in the context of project use cases, included. how would personalised dialogues work inside the chatbot interface, and how could critical thinking abilities be easily baselined and re-assessed. A set of annotated dialogues were reviewed across six Socratic steps by partners to see what the outcomes of a critical thinking approach could be. The six steps are:
Clarification: What do you mean when you say X?
Challenging assumptions: What assumptions are we making here?
Evidence and reasoning: Can you provide an example that supports what you are saying?
Alternative viewpoints: Are there alternative viewpoints?
Implications and consequences: How would this affect someone?
Challenging the question: What do you think was important about that question?
A first glimpse at draft wireframes of what the intelligent coaching agent could look like were shared, enabling deeper discussions and understanding of how to approach experiential learning within solution development. A session on AI trust helped determine how to determine and manage risk issues. All important elements to be taken into the Living Labs which will initially explore critical thinking before moving into dialogue testing to inform design requirements.
After the meeting, Massimo Magaldi from Engineering SpA, the project manager of TITAN , said: “TITAN is pushing the boundaries of innovation, covering many complex topics from critical thinking, AI trust, to Socratic dialogues and disinformation signals. The face-to-face meetings in Rome were really valuable and allowed us to dive into solution details with the other partners.”
“ We are particularly pleased with the level of engagement and progress with the co-design. We aim to have more interactive wireframes of our solution for specific user cases to share and gain feedback from end-users soon.”
For updates on TITAN's progress subscribe to newsletter updates in the footer of www.titanthinking.eu