Confirmation bias

How to prevent AI from only confirming what you already think?

1250 € per session, excl. VAT

In short

Improve decision-making by challenging your assumptions with AI. This course teaches you to identify and counteract your own cognitive biases, such as confirmation bias, when interacting with AI assistants. Learn practical techniques to solicit diverse, critical, and alternative perspectives from AI, leading to more robust strategic insights.

Why this training?

AI assistants are designed to be helpful, but this can lead them to agree with your stated premises, creating an intellectual echo chamber. This phenomenon, known as sycophantic agreement, reinforces your existing beliefs rather than challenging them. This course addresses the human side of the interaction, focusing on how our own cognitive biases, like confirmation bias, shape our prompts and, consequently, the AI's answers. By learning to recognise and counteract these tendencies, you can transform your AI assistant from a simple validator into a powerful tool for critical thinking and more robust decision-making. Participants will work on practical, anonymised business scenarios. They will be tasked with drafting initial prompts to solve a problem, then identifying the inherent biases within those prompts. Subsequently, they will apply techniques like opposite framing and asking for disconfirming evidence to generate a wider range of outputs from the AI. The goal is to compare the initial, biased output with the more diverse, critically-informed results, highlighting the value of these techniques in a tangible way. Deliverables: A personal checklist for identifying cognitive biases in prompts.; A set of prompt templates for generating counterarguments and alternative scenarios.; A structured framework for conducting a pre-mortem analysis with an AI assistant.. This course focuses on the user's cognitive biases and prompt strategies. It is not a course on fact-checking or verifying the source of AI-generated information, which is a distinct skill set. The techniques learned help broaden perspective but do not guarantee the factual accuracy of AI outputs. The workshop is delivered in person as a private session for up to 12 participants, in one language chosen from French, Dutch and English. All exercise data is fictional or anonymised, and every result produced with AI is reviewed and validated by a responsible person. The aim is a bounded result the team can reuse, not a promise of deployment.

Objectives of this training

  • Identify confirmation bias and framing effects in your own AI prompts.
  • Apply neutral and opposite framing techniques to prompts to elicit different responses.
  • Use pre-mortem analysis techniques with an AI to anticipate potential failures.
  • Structure prompts that explicitly ask for disconfirming evidence and alternative viewpoints.
  • Maintain a decision journal to track and reflect on AI-assisted judgments.

Who is it for and what are the prerequisites?

Roles and tasks concerned

  • Managers and team leaders using AI for strategic analysis and decision support.
  • Analysts and researchers seeking to leverage AI for unbiased exploration of topics.
  • Professionals in any role who rely on AI-generated content for reports, proposals, or creative work.

Prior skills

  • Participants should have completed the 'AI Fundamentals' and 'AI Prompting' courses or possess equivalent experience. The core skill of bias awareness is human-centric and can be practised without any specific tool. For the AI-assisted exercises, participants will need access to an AI assistant authorised by their organisation, which may require a paid plan with specific features, along with real user accounts and permissions. All case studies used are fictional or anonymised.

What you will work on

  1. Recognising Your Own Echo Chamber

    Understanding confirmation bias and its impact on AI interactions. Identifying leading questions and framing effects in your prompts.

  2. Techniques for Neutral and Opposite Framing

    Rewriting prompts to remove biased language and assumptions. The power of asking for the opposite: prompting for counterarguments.

  3. Seeking Disconfirming Evidence and Pre-Mortem Analysis

    How to prompt an AI to act as a 'devil's advocate'. Conducting a pre-mortem: asking an AI to imagine a project has failed and explain why.

  4. Systematising Critical Feedback and Reflection

    Addressing the sycophantic tendencies of some AI models. Using a decision journal to document and learn from AI-assisted choices.

What you take away

  • A personal checklist for identifying cognitive biases in prompts.
  • A set of prompt templates for generating counterarguments and alternative scenarios.
  • A structured framework for conducting a pre-mortem analysis with an AI assistant.

Method and assessment

  • Interactive presentations on the psychology of cognitive biases.
  • Hands-on workshops with individual and group exercises using authorised AI tools.
  • Peer review and discussion of prompt strategies and generated outputs.

Continuous assessment through participation in exercises, quality of prompts developed, and a final debrief where participants share their key learnings and action plans.

Frequently asked questions

What concrete result does this course deliver?

Develop the human skill of bias awareness to unlock more critical, diverse, and valuable insights from any AI assistant. Deliverables: a personal checklist for identifying cognitive biases in prompts., a set of prompt templates for generating counterarguments and alternative scenarios., a structured framework for conducting a pre-mortem analysis with an ai assistant..

Who is this course for and what are the prerequisites?

It is designed for: managers and team leaders using ai for strategic analysis and decision support., analysts and researchers seeking to leverage ai for unbiased exploration of topics., professionals in any role who rely on ai-generated content for reports, proposals, or creative work.. Participants should have completed the 'AI Fundamentals' and 'AI Prompting' courses or possess equivalent experience. The core skill of bias awareness is human-centric and can be practised without any specific tool. For the AI-assisted exercises, participants will need access to an AI assistant authorised by their organisation, which may require a paid plan with specific features, along with real user accounts and permissions. All case studies used are fictional or anonymised.

What data and access need to be arranged?

Participants are requested to come with access to their organisation's approved AI assistant; no other specific data preparation is required. Required tools and licences are checked before the session; no subscription is included in the price.

What practical exercise is completed?

Participants will work on practical, anonymised business scenarios. They will be tasked with drafting initial prompts to solve a problem, then identifying the inherent biases within those prompts. Subsequently, they will apply techniques like opposite framing and asking for disconfirming evidence to generate a wider range of outputs from the AI. The goal is to compare the initial, biased output with the more diverse, critically-informed results, highlighting the value of these techniques in a tangible way.

What are the limits and who validates the results?

This course focuses on the user's cognitive biases and prompt strategies. It is not a course on fact-checking or verifying the source of AI-generated information, which is a distinct skill set. The techniques learned help broaden perspective but do not guarantee the factual accuracy of AI outputs.

Can the course be adapted to our line of work?

The workshop's content and exercises can be tailored to the specific professional context and challenges of the participants, using relevant industry scenarios to ensure the practical application of bias-awareness techniques in their daily work.

How is the session organised and what does it cost?

Half-day, private in-person session for up to 12 participants. The fee is €1,250 excluding VAT per session. Date and location are agreed during scoping; each session runs in a single language: French, Dutch or English.

Last updated: