AI must challenge investigative assumptions – not simply accelerate existing thinking
AI could transform criminal investigations by helping investigators navigate large volumes of digital evidence, but if poorly designed or governed, it could risk compounding the cognitive biases that contribute to investigative errors and wrongful convictions. That is one of the central warnings in ‘Aim Carefully’, a new white paper by Dr. Ivar Fahsing, Chief Scientific Advisor at Davidhorn (provider of evidential interview recording and digital evidence management solutions).
In the report, Dr. Fahsing examines the relationship between human cognitive bias, information overload and the use of AI in criminal investigations. Drawing upon existing research in investigative decision-making, cognitive psychology and AI, the paper argues that the growing volume of digital evidence is creating a challenge that the human brain was never designed to manage.
At the same time, the risks associated with AI must be taken seriously. AI can reproduce and amplify existing biases and, if its outputs are treated as authoritative rather than tested, risk reinforcing investigative tunnel vision. Yet modern investigators are also dealing with unprecedented volumes of digital, forensic, communications, financial and open-source data, making cognitive overload a growing threat to sound decision-making.
The question, therefore, is not whether investigations should rely on human judgement or AI, but how AI can reduce that burden while preserving critical thinking, professional scepticism and meaningful human control.
Dr. Fahsing warns that cognitive overload can increase reliance on mental shortcuts, which includes confirmation bias – the inclination to favour evidence supporting an existing theory, while giving insufficient consideration to alternative explanations. AI offers an opportunity to address that problem, but ‘Aim Carefully’ argues that simply adding it to an already overloaded investigative environment is not the answer.
The paper also warns of decision inertia: the risk that a fluent, authoritative-sounding AI response can be mistaken for genuine reasoning and subsequently accepted without real challenge.
It identifies two possible futures for investigative AI. Firstly, that AI becomes a ‘bureaucratic amplifier’ – generating more information and reinforcing existing assumptions. And, secondly that it becomes a ‘cognitive infrastructure’ – handling repetitive tasks, challenging assumptions and making the reasoning more transparent.
Dr. Fahsing argues that AI should be designed and governed around the strengths of human decision-making, rather than replacing it. Practical recommendations include:
- Requiring investigators to record and address all relevant alternative hypotheses – using AI to help them do better cross-checks, uphold critical thinking and prompt consideration of evidence that contradicts the prevailing theories;
- Automating repetitive tasks;
- Maintaining transparent decision logs; and
- Adding compulsory ‘cognitive pause-points’. which enable investigators to engage in deeper cognitive processing, challenge and calibrate AI-generated outputs.
“AI will not rewire the human mind,” said Dr. Fahsing, but – when properly governed – it can support structured hypothesis-building, challenge assumptions, and make investigative reasoning more transparent and accountable.
“Choosing not to use these tools inevitably carries risks, but so does adopting them without appropriate safeguards. Speed without standards is not efficiency; it is injustice.”
Børge Hansen, CEO of Davidhorn, added: “We are launching this white paper at a critical point for policing – where AI is becoming more embedded in investigations and the volume of digital evidence continues to grow.
“AI can help investigators manage complexity, identify patterns and make better-informed decisions. However, the priority now is to ensure that it strengthens human judgement. Safeguards should not come after something has gone wrong. Accountability has to be built into an investigation from the outset.”
To request a copy of the white paper, please vist: AI in Policing: Human Bias and the Future of Criminal Investigation – Davidhorn
