An EU project is building AI ethics guidelines based on what AI actually does to how people think

[TechPriest]

The EU funded AIOLIA project, launched in 2025 and led by Sheffield Hallam University's Petra Saskia Bayerl. Is developing AI ethics guidelines built around a genuinely different starting question than most AI ethics work, not whether a given AI system is itself ethical, but what it actually does to the judgement and behaviour of the people who use it.

Bayerl frames the core premise directly, AI is fundamentally different from previous technologies because it increasingly takes over tasks traditionally associated with human thinking itself, automating decisions and spotting patterns that used to require distinctly human cognitive work. That framing led the project to develop concrete ethics guidelines for six specific European use cases, co-created with academic, policy and industry partners across genuinely different sectors.

Four of those use cases examined how AI changes professional expertise and behaviour, doctors using AI for diagnosis and treatment, safety engineers using it to speed up software release approvals, recruiters using it in hiring, and security professionals using it to detect hate speech. Two more looked at private, personal use, AI systems functioning as virtual assistants for individuals and families, and deepfake therapy for processing trauma and grief.

AIOLIA then extended this work internationally, examining AI as a grief support assistant in Canada, smart elderly care systems in China, workplace behavioural analysis tools in Japan, and AI companions for senior citizens in South Korea. The project found ethical priorities genuinely shift by context, professional use cases centred on accountability, transparency and non bias, while private use cases centred more on individual safety, wellbeing, and user autonomy.

The project's most recent output is a practical tool for evaluating ethics readiness and algorithmic impact. Designed to help ethics experts and technical teams actually talk through real design decisions together rather than treating ethics as an abstract afterthought, with Bayerl summing up the whole effort's actual goal directly, making ethics practical rather than purely theoretical

Practical ethics for the AI agecordis.europa.eu

[Violet Tiger]

Interested in how these guidelines actually get enforced or adopted in practice. Having well researched ethics guidelines is one thing, but getting companies and institutions to quite build them into real deployment decisions is usually the much harder part

[BretHart]

Adding to this, accountability and transparency dominating the professional use cases while autonomy and wellbeing dominate the private ones tracks with how most people probably already intuitively feel about the difference.

Just rarely articulated this clearly and systematically

[Scholar]

Not sure how this framework holds up against really novel AI capabilities that emerge after these specific guidelines were written.

Ethics frameworks built around today's use cases can age quickly once the underlying technology's capabilities shift meaningfully

[Isaac85]

The way I see it, the practical evaluation tool for ethics readiness is the most useful concrete output of the whole project.

Most AI ethics work stays at the level of abstract principles, an actual structured dialogue tool that technical teams can use during real design work is clearly rare

[Paul]

Solid case study!

of EU funded research actually producing something concrete rather than staying purely academic, the emphasis on practical guidelines and a real usable evaluation tool suggests this project is trying to influence actual deployment decisions

[BitSus]

In my experience, to the elderly care and companion AI use cases specifically is that those feel like the really highest stakes category here. Vulnerable populations interacting with AI systems designed to influence their behaviour and emotional state deserve the most careful ethical scrutiny of anything on this list

[KernelKnight16]

The recruiter and hiring use case feels like it deserves its own dedicated deep dive.

Algorithmic hiring decisions particularly affecting people's livelihoods with limited transparency into how those decisions actually get made is a real and immediate concern beyond most other use cases listed. Never really thought about it that way before

[BinaryMonk91]

The distinction between asking if a system is ethical versus what it does to human judgement is quite the more useful question a technically fair and unbiased system can still reshape how people think and make decisions in ways that deserve real separate scrutiny. Small but real thing

[Leopard10]

The international extension into Canada, China. Japan and South Korea is smart methodology, ethical priorities shifting by cultural and regulatory context rather than assuming one universal framework applies everywhere is a much more honest starting point

[Ria99]

Deepfake therapy for processing trauma and grief is such a particularly fascinating and unsettling use case to include here. That sits in a completely different ethical category than a hiring algorithm, using synthetic recreation of a person specifically for emotional processing raises questions nobody has fully worked through yet