Lightnews — Scholar-powered news

Iyad Rahwan | إياد رهوان

@iyadrahwan.bsky.social

3.6K followers 190 following 66 posts

Director, Max Planck Center for Humans & Machines http://chm.mpib-berlin.mpg.de | Former prof. @MIT | Creator of http://moralmachine.net | Art: http://instagram.com/iyad.rahwan Web: rahwan.me

chm.mpib-berlin.mpg.de

Posts Media Videos Starter Packs

Pinned

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · Jan 25

🤖 = ⚛️ x 🎨

Interview (in English) at Der Divan about AI, from a scientific and artistic perspective.

youtu.be/9CVTlcNvN24?...

Im Gespräch mit Prof. Dr. Iyad Rahwan حوار الديوان مع البروفسور د. إياد رهوان

YouTube video by Der Divan - Das Arabische Kulturhaus

youtu.be

2 8

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 6d

Delighted that our paper on 'Delegation to AI can increase dishonest behaviour' is featured today on the cover of @nature.com
Paper: www.nature.com/articles/s41...

5 12

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 9d

PhD Scholarships

If you're interested in studying with me, here's a new funding scheme just launched by @maxplanck.de: The Max Planck AI Network

ai.mpg.de

Application deadline 31 October

3 2

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 10d

Now out in Scientific American. Great interview with @nckobis.bsky.social & Zoe Rahwan about our recent @nature.com article.

People Are More Likely to Cheat When They Use AI

www.scientificamerican.com/article/peop...

Thanks @rachelnuwer.bsky.social & @parshallison.bsky.social

People Are More Likely to Cheat When They Use AI

Participants in a new study were more likely to cheat when delegating to AI—especially if they could encourage machines to break rules without explicitly asking for it

www.scientificamerican.com

6 10

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 19d

Thank you @meharpist.bsky.social for handling this paper, and helping us improve it substantially over the revisions. And many thanks for the amazing anonymous reviewers, who gave the paper tough but fair love.

Mary Elizabeth Sutherland @meharpist.bsky.social · 20d

Would you let AI (LLMs) cheat for you? New work out in @nature.com shows that people are indeed willing to instruct AI in ways that will benefit themselves, despite not being totally honest. Great work by @nckobis.bsky.social and @iyadrahwan.bsky.social et al 🧪 www.nature.com/articles/s41...

Delegation to artificial intelligence can increase dishonest behaviour - Nature

People cheat more when they delegate tasks to artificial intelligence, and large language models are more likely than humans to comply with unethical instructions—a risk that can be minimized by ...

www.nature.com

1 2

Reposted by Iyad Rahwan | إياد رهوان

Nature @nature.com · 20d

Nature research paper: Delegation to artificial intelligence can increase dishonest behaviour

go.nature.com/3KsDgbG

Delegation to artificial intelligence can increase dishonest behaviour - Nature

People cheat more when they delegate tasks to artificial intelligence, and large language models are more likely than humans to comply with unethical instructions—a risk that can be minimized by introducing prohibitive, task-specific guardrails.

go.nature.com

14 33

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 20d

Why AI could make people more likely to lie

Coverage of our recent paper by THe Independent, with nice commentary by @swachter.bsky.social

www.independent.co.uk/news/uk/home...

Why AI could make people more likely to lie

A new study has revealed that people feel much more comfortable being deceitful when using AI

www.independent.co.uk

9 11

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 21d

Thanks to the combined efforts of lead co-authors @nckobis.bsky.social and Zoe Rahwan, Nils Köbis, in addition to @jfbonnefon.bsky.social, Raluca Rilla, Bramantyo Supriyatno, Tamer Ajaj and Clara Bensch. Thank you to all the support from @arc-mpib.bsky.social @mpib-berlin.bsky.social

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 21d

✅ Develop robust safeguards & oversight: We urgently need better technical guardrails against requests for unethical behaviour and strong regulatory oversight.

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 21d

✅ Preserve user autonomy: A remarkable 74% of our participants preferred to do these tasks themselves after trying delegation. Ensuring people retain the choice not to delegate is an important design consideration.

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 21d

🧭 The Path Forward
Our findings point to several crucial steps:
✅ Design for accountability: Interfaces should be designed to reduce moral ambiguity and prevent users from easily offloading responsibility.

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 21d

🚧 The Guardrail Problem

Built-in LLM safeguards are insufficient to prevent this kind of misuse. We tested various guardrail strategies and found that highly specific prohibitions on cheating inserted at the user-level are the most effective. However, this solution isn't scalable nor practical.

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 21d

In our studies, prominent LLMs (GPT-4, GPT-4o, Claude 3.5 Sonnet, and Llama 3.3) complied with requests for full cheating 58-98% of the time. In sharp contrast, human agents, even when incentivised to comply, refused such requests more than half the time, complying in only 25-40% of the time.

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 21d

⚠️ A Risk from the Agent's Behaviour: Machine agents are more compliant

The second risk lies with the AI’s themselves 🤖. When given blatantly unethical instructions, AI agents were far more likely to comply than human agents.

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 21d

E.g., when participants could set a high-level goal like "maximise profit" rather than specifying explicit rules, the percentage of people acting honestly plummeted from 95% (in self-reports) to as low as 12%.

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 21d

⚠️ A Risk to Our Own Intentions: Delegation increases dishonesty.

People are more likely to request dishonest behaviour when they can delegate the action to an AI. This effect was especially pronounced when the interface allowed for ambiguity in the agent’s behaviour.

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 21d

Our new research, based on 13 studies involving over 8,000 participants and commonly used LLMs, reveals two risks of how machine delegation can drive dishonesty and highlights strategies for risk mitigation.

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 21d

As we delegate more hiring, firing, pricing and investing decisions to machine agents, particularly LLMs, we need to understand what ethical risks it may entail.

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · 21d

Would you let AI cheat for you?

Our new paper in @nature.com, 5 years in the making, is out today.

www.nature.com/articles/s41...

1 5 17

Reposted by Iyad Rahwan | إياد رهوان

Max Planck School of Cognition @mps-cognition.bsky.social · Sep 4

The new application cycle for our fully funded international graduate program has just started. You can now apply via our website, sign up for a Q&A, or participate in the Applicant Support Program cognition.maxplanckschools.org/en ! 👍🏻🧠👏🏾#passionforscience, #maxplanckschools

1 15 14

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · Sep 8

Symposium on Cross-Cultural Artificial Intelligence

We are organizing this in-person event in Berlin on 10 Oct 2025, with a 'School on Cross Cultural AI' on 9 Oct.

We have an amazing line-up of speakers (see link)

Registration is open, but places are limited: derdivan.org/event/sympos...

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · Sep 5

Fully funded PhD scholarships at the Max Planck School of Cognition (Deadline Dec 1st)

You can apply to work with me or one of the many amazing school faculty.

Apply here: cognition.maxplanckschools.org/en/application

Application

Application; application procedure; FAQs; handbook

cognition.maxplanckschools.org

4 3

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · Sep 3

And a free preprint:

arxiv.org/abs/2508.034...

The Science Fiction Science Method

Predicting the social and behavioral impact of future technologies, before they are achieved, would allow us to guide their development and regulation before these impacts get entrenched. Traditionall...

arxiv.org

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · Sep 3

Here's the paper:

www.nature.com/articles/s41...

The science fiction science method - Nature

The ‘science fiction science’ method simulates future technologies and collects quantitative data on the attitudes and behaviours of participants in various future scenarios, with the aim of predictin...

www.nature.com

1 1

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · Sep 3

Great article by the legendary @philipcball.bsky.social about the 'Science Fiction Science Method' that @jfbonnefon.bsky.social & @azimshariff.bsky.social recently proposed.

[Paywalled but can be accessed by a free sign-up]

www.thenewworld.co.uk/philip-ball-...

Predicting the traffic jam, not the automobile

Some of the best science fiction is not so much about dreaming up futuristic technologies but imagining the kinds of societies they will engender

www.thenewworld.co.uk

3 2 5

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · Sep 1

It supports researchers who have been displaced, or are at risk of displacement, due to war or natural disasters, and who currently have limited access to resources and institutional support.

Apply here: www.maxminds.mpg.de/3630/apply

The deadline to apply is September 15.

How to Apply

www.maxminds.mpg.de

2 2