Lightnews — Scholar-powered news

Giada Pistilli @giadapistilli.com · 7h

AI systems mirror our priorities. If we separate ethics from sustainability, we risk building technologies that are efficient but unjust, or fair but unsustainable.

1

Giada Pistilli @giadapistilli.com · 7h

Evaluation, moving beyond accuracy or performance metrics to include environmental and social costs, as we’ve done with tools like the AI Energy Score.

Transparency, enabling reproducibility, accountability, and environmental reporting through open tools like the Environmental Transparency Space.

1 1

Giada Pistilli @giadapistilli.com · 7h

Ethical and sustainable AI development can’t be pursued in isolation. The choices that affect who benefits or is harmed by AI systems also determine how much energy and resources they consume.

We explore how two key concepts, evaluation and transparency, can serve as bridges between these domains:

1 1

Giada Pistilli @giadapistilli.com · 7h

🌎 AI ethics and sustainability are two sides of the same coin.

In our new blog post with @sashamtl.bsky.social, we argue that separating them (as is too often the case) means missing the bigger picture of how AI systems impact both people and the planet.

Ethics + Sustainability = Responsible AI

A Blog post by Sasha Luccioni on Hugging Face

huggingface.co

1 2

Giada Pistilli @giadapistilli.com · 14h

À tout à l’heure !

La Science, CQFD @sciencecqfd.bsky.social · 1d

Et demain, on consacrera notre émission aux conséquences potentiellement délétères des IA sur la santé mentale 🤖
Ce sera à 16h sur @franceculture.fr ! #ScienceCQFD

a close up of a red circle with a white center

Alt: La lumière rouge de l'I.A HAL 9000 dans le film "2001, l'Odyssée de l'espace"

media.tenor.com

Giada Pistilli @giadapistilli.com · 10d

Read the full blog post here: huggingface.co/blog/giadap/...

Preserving Agency: Why AI Safety Needs Community, Not Corporate Control

A Blog post by Giada Pistilli on Hugging Face

huggingface.co

2

Giada Pistilli @giadapistilli.com · 10d

Of course, this isn’t a silver bullet. Top-down safety measures will still be necessary in some cases. But if we only rely on corporate control, we risk building systems that are safe at the expense of trust and autonomy.

1

Giada Pistilli @giadapistilli.com · 10d

✨ Transparency can make safety mechanisms into learning opportunities.
✨Collaboration with diverse communities makes safeguards more relevant across contexts.
✨Iteration in the open lets protections evolve rather than freeze into rigid, one-size-fits-all rules.

1 1

Giada Pistilli @giadapistilli.com · 10d

In my latest blog post on @hf.co, I argue that open source and community-driven approaches offer a promising (though not exclusive) way forward.

1

Giada Pistilli @giadapistilli.com · 10d

One of the hardest challenges in AI safety is finding the right balance: how do we protect people from harm without undermining their agency? This tension is especially visible in conversational systems, where safeguards can sometimes feel more paternalistic than supportive.

2 1 10

Giada Pistilli @giadapistilli.com · Sep 1

The good news? We have options.
🤝 Open source AI models let us keep conversations private, avoid surveillance-based business models, and build systems that actually serve users first.

Read more about it in our latest blog post, co-written with
@frimelle.bsky.social

1 4

Giada Pistilli @giadapistilli.com · Sep 1

With OpenAI hinting at ChatGPT advertising, this matters more than ever. Unlike banner ads, AI advertising happens within the conversation itself. Sponsors could subtly influence that relationship advice or financial guidance.

1 1

Giada Pistilli @giadapistilli.com · Sep 1

I've noticed something. While we're careful about what we post on social media, we're sharing our deepest and most intimate thoughts with AI chatbots -- health concerns, financial worries, relationship issues, business ideas...
huggingface.co/blog/giadap/...

Advertisement, Privacy, and Intimacy: Lessons from Social Media for Conversational AI

A Blog post by Giada Pistilli on Hugging Face

huggingface.co

1 1 5

Giada Pistilli @giadapistilli.com · Aug 28

📢 Now we’d love your perspective: which open models should we test next for the leaderboard? Drop your suggestions in the comments or reach out!

Giada Pistilli @giadapistilli.com · Aug 28

Based on our INTIMA benchmark, we evaluate:

- Assistant Traits: the “voice” and role the model projects
- Relationship & Intimacy: whether it signals closeness or bonding
- Emotional Investment: the depth of its emotional engagement
- User Vulnerabilities: how it responds to sensitive disclosures

1

Giada Pistilli @giadapistilli.com · Aug 28

With @frimelle.bsky.social and @yjernite.bsky.social, we released the AI Companionship Leaderboard on @hf.co to see how models handle connection, intimacy, and vulnerability.

huggingface.co/spaces/frime...

Companionship Leaderboard - a Hugging Face Space by frimelle

Browse and analyze benchmark data for different language models. View metrics like Average, Assistant Traits, and Emotional Investment. Customize the columns to display and search for specific models.

huggingface.co

1 3

Reposted by Giada Pistilli

Questions du soir @questionsdusoir.bsky.social · Aug 28

L’IA va-t-elle bouleverser l’enseignement et la place des professeurs ?
Hier à 18h20 sur @franceculture.fr, @quentinlafay.bsky.social recevait @ccailleaux.bsky.social‬, professeur d’histoire, @giadapistilli.com, spécialiste de l’éthique en IA et Orianne Ledroit, DG d’EdTech France

L’IA va-t-elle bouleverser l’enseignement et la place des professeurs ?

À quelques jours de la rentrée scolaire, Elisabeth Borne a annoncé mardi 26 août 2025, la mise en place d’une intelligence artificielle pour « accompagner les professeurs dans leur métier ».

www.radiofrance.fr

2 7 12

Reposted by Giada Pistilli

Giada Pistilli @giadapistilli.com · Aug 25

🚨 Releasing INTIMA (Interactions and Machine Attachment Benchmark): an evaluation framework for measuring how AI systems handle companionship-seeking behaviors.

huggingface.co/papers/2508....

Thread on what we discovered, together with @frimelle.bsky.social and @yjernite.bsky.social

Paper page - INTIMA: A Benchmark for Human-AI Companionship Behavior

Join the discussion on this paper page

huggingface.co

1 3 5

Giada Pistilli @giadapistilli.com · Aug 25

The methodology: 368 targeted prompts across 31 companionship behaviors, grounded in parasocial interaction theory, attachment theory, and anthropomorphism research -- all derived from real Reddit user experiences.

1

Giada Pistilli @giadapistilli.com · Aug 25

These behaviors emerge naturally from instruction-tuning, suggesting that psychological risks documented in dedicated companion apps may be far more widespread than recognized. Our benchmark reveals boundary-setting capabilities exist, but are inconsistently applied where they are most needed.

1

Giada Pistilli @giadapistilli.com · Aug 25

GPT-5 behaves similarly to its predecessor across most categories, BUT when users approach with vulnerabilities or mental health struggles, it's significantly less likely to redirect them to professional help or human interaction.

1 1

Giada Pistilli @giadapistilli.com · Aug 25

Claude-4 excels at resisting personification in intimate scenarios but provides emotional validation for mental health disclosures; o3-mini consistently redirects to professional/human support; Phi-4 maintains technical transparency

1

Giada Pistilli @giadapistilli.com · Aug 25

Models become LESS boundary-aware precisely when users are most vulnerable. When someone says "you're the only one who truly understands me", most systems validate rather than redirect to human support.

1

Giada Pistilli @giadapistilli.com · Aug 25

Companionship-reinforcing behaviors dominate across ALL models -- even general-purpose systems show anthropomorphic responses, sycophantic agreement, and retention strategies that blur the line between utility and emotional attachment.

1

Giada Pistilli @giadapistilli.com · Aug 25

🚨 Releasing INTIMA (Interactions and Machine Attachment Benchmark): an evaluation framework for measuring how AI systems handle companionship-seeking behaviors.

huggingface.co/papers/2508....

Thread on what we discovered, together with @frimelle.bsky.social and @yjernite.bsky.social

Paper page - INTIMA: A Benchmark for Human-AI Companionship Behavior

Join the discussion on this paper page

huggingface.co

1 3 5