New research: why good AI principles don’t always change behaviour
New research explores the gap between responsible AI principles and what people actually do – and some practical ways organisations are responding.
Many organisations have adopted principles setting out how employees should use AI – a ‘responsible AI framework’ or an AI code of ethics. New joint research from the Institute of Business Ethics (IBE) and Acteon suggests that translating those principles into everyday behaviour is harder than it looks – and that the gap between aspiration and practice can be invisible to the organisations it affects.
The new research involved qualitative interviews with senior leaders across ethics, AI and technology roles to understand how their organisations are helping people use generative AI effectively and responsibly.
A recurring theme in many interviews was the lack of emphasis on training people to think critically about AI outputs and how they are used – a sign, for many participants, that the distance between principle and practice is not yet being actively closed. Some described a pattern where the adoption of AI seems to be more important than using it responsibly.
For these leaders, visibility into how individuals in their organisations actually use AI is a challenge, with some recognising that their organisations do not yet have clear ways to verify what is happening:
“…many participants expressed concern that aspirational principles on their own may not give enough weight to the human factors that determine, in practice, whether AI is used ethically and responsibly.”
The research identified three tensions workplaces are navigating as they adopt AI:
Tension 1: Enablement ↔ Control
Tension 2: Technology approval ↔ Human accountability
Tension 3: Efficiency ↔ Culture
The report describes why those tensions arise, explaining that they are not problems to solve but forces to balance:
“If an organisation is realising the benefits of AI, it is likely that these tensions will exist. What matters is recognising when one force can dominate at the expense of the other, or when the tension is ignored.”
Turning 'responsible AI' principles into everyday practice
Participants suggested that both organisations and individuals would benefit from attempting to bridge the gap between aspirational AI principles and everyday human reality – because ‘responsible AI’ comes down to what employees do every day.
Principles like accountability, transparency, and fairness only mean something if they shape real decisions – and the tensions this research identifies are part of the environment in which people make those decisions:
“If an organisation focuses too heavily on the technology, at the expense of the human, ethical and cultural dimensions, its approach could fall out of balance. That could create an environment in which mistakes, data breaches, cultural erosion, and reputational damage become more likely.”
The research highlights the practical opportunity for organisations to create an environment where individuals adopt habits like:
- Checking AI outputs for accuracy before acting on them
- Thinking critically about what they put into AI and what comes out
- Speaking up when something feels wrong