This lesson plan sets out how journalists can use artificial intelligence (AI) in investigative journalism.
It’s based on the article AI and investigative journalism which we suggest you read before adapting for your own purposes.
Introduction
The integration of Artificial Intelligence (AI) into investigative journalism offers powerful opportunities to uncover stories hidden within massive datasets, automate tedious tasks, and enhance visual storytelling. However, it also introduces significant ethical risks, including bias, lack of transparency, and the potential for misinformation. This day-long lesson is designed to equip journalists and media students with a practical and ethical framework for using AI tools responsibly in their investigations.
Sessions timetable
09:00–10:00 – Session 1: Introduction to AI in investigative reporting
Aims: To define AI within a journalistic context and identify its primary benefits and risks.
Presentation: Define terms such as Machine Learning (ML), Large Language Models (LLM), and Natural Language Processing (NLP). Using the MHM content, explain how AI acts as an assistant rather than a replacement for human judgment. Highlight the three pillars: efficiency, discovery, and verification.
Activity: Participants work in pairs to list three things they think AI can do and three things it cannot. Compare these against the realities of current technology.
Discussion: Why must the journalist remain in the loop? Discuss the dangers of over-reliance on automated tools.
10:00–11:00 – Session 2: Data scraping and pattern recognition
Aims: To understand how AI can be used to process large volumes of unorganised data.
Presentation: Demonstrate how AI tools can extract information from thousands of PDFs or spreadsheets. Discuss pattern recognition – identifying anomalies in financial records or satellite imagery that might indicate corruption or environmental crimes.
Activity: Data categorisation exercise. Provide trainees with a sample list of 100 diverse public records and ask them to brainstorm how an AI tool could categorise these faster than a human.
Discussion: If an AI finds a pattern, how do we prove it is real and not a hallucination or a statistical fluke?
11:00–11:15 – Break
11:15–12:45 – Session 3: Open-source intelligence (OSINT) and AI
Aims: To explore how AI enhances investigative techniques using publicly available information.
Presentation: Cover the use of AI in geolocation, facial recognition (and its ethical pitfalls), and the analysis of social media trends. Explain how AI can help track ships, planes, or deforestation patterns through satellite data.
Activity: Geolocation challenge. Using a provided image, trainees use AI-enhanced image search tools and maps to identify the specific location and time the photo was taken.
Discussion: The ethics of surveillance. When does using AI to track individuals cross the line from investigative journalism to an invasion of privacy?
12:45–13:45 – Lunch
13:45–15:00 – Session 4: Ethics, bias, and the hallucination risk
Aims: To critically evaluate the reliability of AI-generated outputs and the inherent biases in training data.
Presentation: Focus on the MHM material regarding ethical considerations. Explain how algorithms can inherit the prejudices of their creators. Discuss hallucinations – where AI confidently presents false information as fact.
Activity: Provide an AI-generated summary of a complex legal case containing three subtle factual errors. Trainees must find the errors using traditional primary sources.
Discussion: Should we tell our audience when we use AI to help find a story? How do we credit the tool?
15:00–15:15 – Break
15:15–16:15 – Session 5: Verifying AI-generated content and deepfakes
Aims: To learn how to identify manipulated media and protect the integrity of the newsroom.
Presentation: Showcase examples of AI-generated images, audio, and video (deepfakes). Use MHM guidance to set out a verification checklist (e.g., checking metadata, looking for visual artefacts, verifying the source).
Activity: Show a series of images and audio clips; trainees must vote on which are authentic and which are synthetic, explaining their reasoning based on specific indicators.
Discussion: The impact of deepfakes on public trust. How can journalists act as the ultimate filter for truth?
16:15–17:00 – Session 6: Building an AI-assisted investigation workflow
Aims: To combine the day’s learning into a practical, step-by-step investigative plan.
Presentation: Summarise how to integrate AI at different stages: Pitching (research), Investigation (data), and Production (visualisation). Emphasise the MHM rule: Verify everything.
Activity: Small groups are given a brief tip-off (e.g., suspicious government spending). They must map out a five-step investigation plan, identifying which AI tools could help and where human verification is mandatory.
Discussion: Group presentations of workflows. Peer review focusing on the ‘safety’ and ‘ethics’ of each plan.
Assignment
The AI Audit: Select a previous investigative story (either your own or a well-known example). Write a 500-word report outlining:
- Two ways AI could have speeded up the investigation.
- One way AI could have discovered an angle that was missed.
- Three specific verification steps you would take to ensures the AI-assisted findings were 100% accurate before publication.
Materials needed
- Laptops with internet access.
- Access to LLM tools (e.g., ChatGPT, Claude) for testing.
- Sample datasets (CSV files) and folders of leaked PDF documents for exercises.
- Projector for displaying examples of deepfakes and data visualisations.
Assessment
- Participation: Active engagement in the exercises.
- Critical thinking: Ability to identify ethical risks and biases during group discussions.
- Methodology: The quality and rigour of the investigative workflow developed in Session 6.
- Assignment Quality: Demonstration of a clear understanding of the human-in-the-loop principle in the final written report.
Summary
This lesson plan provides a comprehensive framework for media trainers to teach the intersection of technology and accountability. By following this structure, trainees will learn to harness the power of AI to handle ‘big data’ while maintaining the rigorous ethical standards and verification protocols essential to investigative journalism.
Original source: AI and investigative journalism – Media Helping Media
Related material
