
The Socratic AI Challenge
Question better, use AI judiciously, decide more clearly.
In short. The Socratic AI Challenge is a serious game (management) for 6–20 participants, lasting 60–105 min. Created by serious-games.com. Use it for: prise de décision, communication, apprentissage.
Card by serious-games.com, updated 22 November 2026. Original game by serious-games.com.
A workshop game where teams investigate a complex professional dilemma by combining Socratic dialogue and AI assistance. Participants earn points not by being right, but by formulating better questions, testing their assumptions, and improving their reasoning. The game trains critical thinking in response to AI-generated answers and demonstrates how to keep human judgment at the centre. It is suitable for training, management, innovation, or transformation contexts.
Walkthrough
- 1
Game framework and common dilemma
10 minThe facilitator welcomes the group, forms teams of 3 to 5, and displays the chosen dilemma, for example: 'Should we automate part of customer support with an internal AI?'. They say: 'Your goal is not to defend a fixed opinion, but to produce the strongest reasoning possible.' Each team receives a score sheet and chooses a spokesperson. This step creates common ground and establishes the central rule: question before concluding.
Tip : Prepare 2 or 3 dilemmas suitable for the audience and let the group vote quickly if the proposed topic does not resonate.
- 2
Mini-training in Socratic questioning
12 minThe facilitator presents 5 families of questions: clarification, evidence, assumptions, consequences, alternative perspectives. They give an oral example: 'What do you mean by effective?', then ask teams to write 2 questions from each family on cards. Participants then read one question aloud. This step builds a common language and prevents the workshop from becoming a simple debate of opinions.
Tip : Value short, open, and neutral questions. Politely decline questions that already contain the answer.
- 3
Initial reasoning without AI
15 minEach team discusses the dilemma without using AI tools and fills out an Argument card: provisional position, 3 reasons, 2 risks, 2 missing pieces of information. The facilitator circulates and says: 'Note what you assume, even if it seems obvious.' Teams score 1 point for each explicit assumption and 1 point for each relevant missing piece of information, with a maximum of 6 points. The aim is to establish a visible human basis before any external influence.
Tip : If a team is too quick to seek the right answer, ask: 'What would make you change your mind?'
- 4
Interrogation of the AI
18 minTeams use AI with a set instruction: only ask questions, objections, or identify blind spots, not a final decision. The facilitator provides a sample prompt: 'Act as a Socratic questioner. Ask 10 questions to test our reasoning, without concluding for us.' Participants select the 5 most useful questions and copy them onto their sheet. They earn 1 point for each question that reveals an assumption, a risk, or a new perspective, with a maximum of 5 points.
Tip : Remind them not to input sensitive information. If the AI provides affirmative answers, ask teams to rephrase to obtain questions.
- 5
Cross-dialogue between teams
20 minTeams pair up. Team A presents its provisional position in 3 minutes, then Team B can only respond with questions for 5 minutes. Roles are then reversed. The facilitator says: 'Your mission is to help the other team think more deeply, not to trap them.' Each team notes the 3 questions received that most compel them to revise their reasoning. This step introduces human contradiction and the quality of listening.
Tip : Intervene if questions become accusatory. Rephrase in a neutral version: 'What evidence supports this idea?'
- 6
Critical revision with human arbitration
15 minEach team revisits its Argument card and produces a revised version: recommended decision, success conditions, points of uncertainty, next test to conduct. The facilitator says: 'AI can open avenues, but you are responsible for what you retain.' Teams assign a status to each AI input: accepted, rejected, or to be verified. They can earn up to 6 points: 2 for a nuanced decision, 2 for explicit uncertainties, 2 for a concrete test.
Tip : Ask for a test feasible within two fictitious weeks: interview, prototype, analysis, experimentation, or peer review.
- 7
Final pitch and robustness vote
15 minEach team presents its final reasoning in 2 minutes: decision, main questions that transformed it, and retained or rejected use of AI. Other teams vote with 3 tokens: robustness of reasoning, quality of questions, discernment in facing AI. The facilitator sums up the game points and votes, then congratulates the most robust team, not the most persuasive. This step anchors the fact that the quality of a decision depends on its critical examination.
Tip : Strictly time the pitches. The short format forces a distinction between the essential and the noise.
- 8
Transfer and commitment to use
10 minThe facilitator invites everyone to write a transferable practice: a question to use in a meeting, a rule for using AI, or a verification reflex. They say: 'Choose an action you can apply at your next collective decision.' Participants share in pairs, then a few volunteers read their commitment. This conclusion transforms the playful experience into an observable professional habit.
Tip : Photograph the collective commitments or transcribe them into the minutes to facilitate follow-up.
Variants
- Short version 60 minutes: remove the cross-dialogue, limit the AI interrogation to 10 minutes, and make pitches 90 seconds long.
- Version without AI tool: replace AI with a prepared Questions card deck provided by the facilitator, drawn at random by the teams.
- Remote version: use breakout rooms, a collaborative board, and a shared document per team; votes are made with dots or a poll.
- Advanced version: impose two roles per team, the Advocate of assumptions and the Guardian of AI discernment, who must validate the final presentation.
Debrief guide
- Which questions most evolved your reasoning, and why?
- When did AI help you think more broadly? When could it have misled you?
- What did you learn about your own initial assumptions?
- How do you distinguish a convincing answer from genuinely robust reasoning?
- What rules for using AI should you adopt in your team decisions?
- How can you integrate more Socratic questioning into your meetings or project reviews?
Facilitator reviews
No review yet. Have you run this game? Your feedback will help the next facilitators.
Similar games
Same objectives or same category, to vary or follow up.

The Beer Game
A supply chain simulation to experience the bullwhip effect and understand the impact of delays.

AI or Intuition?
Make sensitive decisions by balancing evidence, algorithms, and human judgement.

Passing the Baton
A game to make handovers work before hidden traps emerge.

Data Detective
Investigate a 30% drop and make decisions without being fooled by the data.

Fishbowl
Facilitating dialogue in a large group without chaos, through a small circle that speaks while others listen.

Ugli Oranges
A quick negotiation to reveal the difference between defending a position and exploring true interests.