Iyad Rahwan | إياد رهوان
@iyadrahwan.bsky.social
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
A word of gratitude to the anonymous reviewers, the unsung heroes of science.
We recently had the great fortune to publish in @nature.com. We even made the cover of the issue, with a witty tagline that summarizes the paper: "Cheat Code: Delegating to AI can encourage dishonest behaviour"
🧵 1/n
We recently had the great fortune to publish in @nature.com. We even made the cover of the issue, with a witty tagline that summarizes the paper: "Cheat Code: Delegating to AI can encourage dishonest behaviour"
🧵 1/n
October 21, 2025 at 3:20 PM
A word of gratitude to the anonymous reviewers, the unsung heroes of science.
We recently had the great fortune to publish in @nature.com. We even made the cover of the issue, with a witty tagline that summarizes the paper: "Cheat Code: Delegating to AI can encourage dishonest behaviour"
🧵 1/n
We recently had the great fortune to publish in @nature.com. We even made the cover of the issue, with a witty tagline that summarizes the paper: "Cheat Code: Delegating to AI can encourage dishonest behaviour"
🧵 1/n
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...
Paper: www.nature.com/articles/s41...
October 2, 2025 at 7:51 AM
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...
Paper: www.nature.com/articles/s41...
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
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
September 29, 2025 at 11:56 AM
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
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
🚧 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.
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.
September 17, 2025 at 3:53 PM
🚧 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.
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.
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.
September 17, 2025 at 3:53 PM
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.
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%.
September 17, 2025 at 3:53 PM
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%.
⚠️ 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.
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.
September 17, 2025 at 3:53 PM
⚠️ 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.
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.
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...
Our new paper in @nature.com, 5 years in the making, is out today.
www.nature.com/articles/s41...
September 17, 2025 at 3:53 PM
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...
Our new paper in @nature.com, 5 years in the making, is out today.
www.nature.com/articles/s41...
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...
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...
September 8, 2025 at 6:23 PM
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...
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...
If you know a 🚨 Scholar at Risk 🚨 please share!
I am delighted to share that applications are now open for the MAXMINDS mentoring program.
I am delighted to share that applications are now open for the MAXMINDS mentoring program.
September 1, 2025 at 6:05 PM
If you know a 🚨 Scholar at Risk 🚨 please share!
I am delighted to share that applications are now open for the MAXMINDS mentoring program.
I am delighted to share that applications are now open for the MAXMINDS mentoring program.
🚨 New preprint 🚨
Experimental Evidence for the Propagation and Preservation of Machine Discoveries in Human Populations
arxiv.org/abs/2506.17741
with team members @levinbrinkmann.bsky.social @thomasfmueller.bsky.social Ann-Marie Nussberger, @maximederex.bsky.social, Sara Bonati, Valerii Chirkov
Experimental Evidence for the Propagation and Preservation of Machine Discoveries in Human Populations
arxiv.org/abs/2506.17741
with team members @levinbrinkmann.bsky.social @thomasfmueller.bsky.social Ann-Marie Nussberger, @maximederex.bsky.social, Sara Bonati, Valerii Chirkov
June 27, 2025 at 10:33 AM
🚨 New preprint 🚨
Experimental Evidence for the Propagation and Preservation of Machine Discoveries in Human Populations
arxiv.org/abs/2506.17741
with team members @levinbrinkmann.bsky.social @thomasfmueller.bsky.social Ann-Marie Nussberger, @maximederex.bsky.social, Sara Bonati, Valerii Chirkov
Experimental Evidence for the Propagation and Preservation of Machine Discoveries in Human Populations
arxiv.org/abs/2506.17741
with team members @levinbrinkmann.bsky.social @thomasfmueller.bsky.social Ann-Marie Nussberger, @maximederex.bsky.social, Sara Bonati, Valerii Chirkov
👻 Spook the Machine 🤖
AI scare you, but can you scare it?
Check out our new Halloween AI project. Each AI has a different phobia. You can try to spook the machine by crafting a prompt to generate a spooky image.
We are giving away $ prizes for the best images.
spookthemachine.com
AI scare you, but can you scare it?
Check out our new Halloween AI project. Each AI has a different phobia. You can try to spook the machine by crafting a prompt to generate a spooky image.
We are giving away $ prizes for the best images.
spookthemachine.com
October 28, 2024 at 10:31 AM
👻 Spook the Machine 🤖
AI scare you, but can you scare it?
Check out our new Halloween AI project. Each AI has a different phobia. You can try to spook the machine by crafting a prompt to generate a spooky image.
We are giving away $ prizes for the best images.
spookthemachine.com
AI scare you, but can you scare it?
Check out our new Halloween AI project. Each AI has a different phobia. You can try to spook the machine by crafting a prompt to generate a spooky image.
We are giving away $ prizes for the best images.
spookthemachine.com
New pre-print from our lab:
Mutual benefits of social learning and algorithmic mediation for cumulative culture
Read the paper here: arxiv.org/html/2410.00...
Mutual benefits of social learning and algorithmic mediation for cumulative culture
Read the paper here: arxiv.org/html/2410.00...
October 4, 2024 at 5:11 PM
New pre-print from our lab:
Mutual benefits of social learning and algorithmic mediation for cumulative culture
Read the paper here: arxiv.org/html/2410.00...
Mutual benefits of social learning and algorithmic mediation for cumulative culture
Read the paper here: arxiv.org/html/2410.00...
Is this the moment autonomous vehicles become safer than humans? @Waymo is reporting that it is:
waymo.com/safety/impact/
If true, will humans adopt these cars now that they are safer?
Our work (with @azimshariff.bsky.social & @jfbonnefon.bsky.social) suggests not.
doi.org/10.1016/j.tr...
waymo.com/safety/impact/
If true, will humans adopt these cars now that they are safer?
Our work (with @azimshariff.bsky.social & @jfbonnefon.bsky.social) suggests not.
doi.org/10.1016/j.tr...
September 12, 2024 at 10:59 AM
Is this the moment autonomous vehicles become safer than humans? @Waymo is reporting that it is:
waymo.com/safety/impact/
If true, will humans adopt these cars now that they are safer?
Our work (with @azimshariff.bsky.social & @jfbonnefon.bsky.social) suggests not.
doi.org/10.1016/j.tr...
waymo.com/safety/impact/
If true, will humans adopt these cars now that they are safer?
Our work (with @azimshariff.bsky.social & @jfbonnefon.bsky.social) suggests not.
doi.org/10.1016/j.tr...
🎨 🎨 🎨 Art exhibition announcement 🎨🎨🎨
Friends, you are cordially invited to the opening of my 2nd solo exhibition, titled 'Portraits of the Artificial', and curated by Nadim Samman
Time: 7:00pm, Friday 28 June
Location: Schützallee 27, 14169 Berlin
RSVP link: www.eventbrite.de/e/eroffnung-...
Friends, you are cordially invited to the opening of my 2nd solo exhibition, titled 'Portraits of the Artificial', and curated by Nadim Samman
Time: 7:00pm, Friday 28 June
Location: Schützallee 27, 14169 Berlin
RSVP link: www.eventbrite.de/e/eroffnung-...
June 19, 2024 at 1:04 PM
🎨 🎨 🎨 Art exhibition announcement 🎨🎨🎨
Friends, you are cordially invited to the opening of my 2nd solo exhibition, titled 'Portraits of the Artificial', and curated by Nadim Samman
Time: 7:00pm, Friday 28 June
Location: Schützallee 27, 14169 Berlin
RSVP link: www.eventbrite.de/e/eroffnung-...
Friends, you are cordially invited to the opening of my 2nd solo exhibition, titled 'Portraits of the Artificial', and curated by Nadim Samman
Time: 7:00pm, Friday 28 June
Location: Schützallee 27, 14169 Berlin
RSVP link: www.eventbrite.de/e/eroffnung-...
In response to popular demand, we just opened 50 additional registration slots for the Machine+Behavior conference, to be held in Berlin 29-30 April.
machinebehavior.science
machinebehavior.science
March 15, 2024 at 7:13 PM
In response to popular demand, we just opened 50 additional registration slots for the Machine+Behavior conference, to be held in Berlin 29-30 April.
machinebehavior.science
machinebehavior.science
Art about AI 🤖🎨
Delighted to share this Aesthetica Magazine interview about my oil paintings, in which I give faces to AI algorithms.
aestheticamagazine.com/exploratory-...
Delighted to share this Aesthetica Magazine interview about my oil paintings, in which I give faces to AI algorithms.
aestheticamagazine.com/exploratory-...
March 4, 2024 at 2:39 PM
Art about AI 🤖🎨
Delighted to share this Aesthetica Magazine interview about my oil paintings, in which I give faces to AI algorithms.
aestheticamagazine.com/exploratory-...
Delighted to share this Aesthetica Magazine interview about my oil paintings, in which I give faces to AI algorithms.
aestheticamagazine.com/exploratory-...
New paper:
Network structure shapes the impact of diversity in collective learning
We study the interaction between
1. Task complexity (simple vs multi-peaked)
2. Social network density, and
3. Team's skill diversity
We find interesting & non-trivial interactions
www.nature.com/articles/s41...
Network structure shapes the impact of diversity in collective learning
We study the interaction between
1. Task complexity (simple vs multi-peaked)
2. Social network density, and
3. Team's skill diversity
We find interesting & non-trivial interactions
www.nature.com/articles/s41...
February 12, 2024 at 12:45 PM
New paper:
Network structure shapes the impact of diversity in collective learning
We study the interaction between
1. Task complexity (simple vs multi-peaked)
2. Social network density, and
3. Team's skill diversity
We find interesting & non-trivial interactions
www.nature.com/articles/s41...
Network structure shapes the impact of diversity in collective learning
We study the interaction between
1. Task complexity (simple vs multi-peaked)
2. Social network density, and
3. Team's skill diversity
We find interesting & non-trivial interactions
www.nature.com/articles/s41...
Two days left for submitting your work-in-progress abstracts to the Machine+Behavior conference to be held in Berlin (29-30 April).
Also, check out our stellar list of speakers.
machinebehavior.science
Also, check out our stellar list of speakers.
machinebehavior.science
February 5, 2024 at 11:26 AM
Two days left for submitting your work-in-progress abstracts to the Machine+Behavior conference to be held in Berlin (29-30 April).
Also, check out our stellar list of speakers.
machinebehavior.science
Also, check out our stellar list of speakers.
machinebehavior.science
CFP: Delighted to announce the "Machine+Behavior" conference, to be held in Berlin, 29-30 April 2024. Poster submission is now open (deadline 13 Jan).
My co-chairs @azimshariff.bsky.social, @jfbonnefon.bsky.social and I look forward to seeing you in Berlin!
machinebehavior.science
My co-chairs @azimshariff.bsky.social, @jfbonnefon.bsky.social and I look forward to seeing you in Berlin!
machinebehavior.science
December 20, 2023 at 2:11 PM
CFP: Delighted to announce the "Machine+Behavior" conference, to be held in Berlin, 29-30 April 2024. Poster submission is now open (deadline 13 Jan).
My co-chairs @azimshariff.bsky.social, @jfbonnefon.bsky.social and I look forward to seeing you in Berlin!
machinebehavior.science
My co-chairs @azimshariff.bsky.social, @jfbonnefon.bsky.social and I look forward to seeing you in Berlin!
machinebehavior.science
Our perspective paper "Machine Culture" is out in Nature Human Behavior.
Free access version: rdcu.be/drzoS
Free access version: rdcu.be/drzoS
November 20, 2023 at 9:30 PM
Our perspective paper "Machine Culture" is out in Nature Human Behavior.
Free access version: rdcu.be/drzoS
Free access version: rdcu.be/drzoS