4/

5/ 🗓 Save the date! Sept 29, 2025 | 18:00 CEST | Online (Teams) For more details 👉 https://t.co/mY8sZ0oZt7 Please RT & invite your network 🙌

4/ Why it matters: 🤖 AI systems mirror human bias ⚖️ Fairness = ethics + compliance 🗣️ Language is culture—exclusion erases voices To harness #HybridCollectiveIntelligence, we must design inclusive, bias-aware AI for all.

3/🎙 Speakers and Talks 🔹Taha Yasseri — How Our Workplace Gender Biases Extend to AI Colleagues 🔹Roberta Calegari — From Opportunity to Compliance: Gender and Diversity in the AEQUITAS Framework for Fair AI 🔹E.M. Lewis-Jong — Language, data and cultural identity in AI systems

2/ Why attend? We’ll explore how gender & diversity shape human–AI collaboration, and how to design inclusive, fair systems free from bias. #AIethics #InclusiveAI

🧵1/ Welcome to HACID Webinar #4! Join us Monday, September 29, 2025 at 18:00 CEST, for an insightful discussion on "Gender and Diversity aspects in Hybrid Collective Intelligence." ` Register now: https://t.co/ADyEwScWFi

8/8 Bottom line: #HumanAI collaboration can significantly boost diagnostic accuracy—but more real-world testing is needed before clinical use. #FutureOfMedicine #AIethics

7/8 ⚠️ Limitations: - Only text-based cases were used (no real patients) - No analysis of treatment quality - Unclear acceptance by doctors/patients - Risks of bias & discrimination still exist #EthicalAI #MedAI

6/8 This shows major promise for next-gen clinical decision-support systems that combine human & machine intelligence. But there are important caveats...

5/8 The best outcomes came from mixed teams: 👩‍⚕️ + 🤖 > either alone Even adding one AI to a group of humans—or vice versa—boosted diagnostic accuracy. #CollectiveIntelligence

4/8 Results: ✅ AI ensembles outperformed 85% of human diagnosticians. ✅ But humans outperformed AI in certain cases. 🚨 When AI failed, humans often had the correct answer. #ArtificialIntelligence #Diagnostics

3/8 We tested this with >2,100 medical case vignettes—realistic patient descriptions—and compared diagnoses by: 👩‍⚕️ individual doctors 👩‍⚕️👨‍⚕️👩‍⚕️👨‍⚕️ doctor groups 🤖 top AI models 🤖🤖👨‍⚕️👩‍⚕️ human-AI teams #MedTwitter #AIresearch

2/8 Neither humans nor #LLMs are perfect—but they make different mistakes. That difference is key. When combined, their strengths complement each other. 🤝 #HybridIntelligence

🧵1/8 Our latest study in @PNASNews shows a powerful insight: combining human doctors with #AI (LLMs) leads to more accurate medical diagnoses. 🔗 https://t.co/WVfoayjvv6 #HealthTech #AIinMedicine

Excited to share that HACID is sponsoring the 2025 International Semantic Web Research Summer School! We're funding student grants to support early-career researchers exploring Knowledge Graphs for Reliable AI, a mission at the heart of HACID. Learn more:

The paper “Ontology Generation Using Large Language Models” shows how LLMs can support ontology engineering, outperforming novice modellers with the right prompts! This work is supported by HACID, further contributing to hybrid collective intelligence. https://t.co/f7i1lh9ltB

Last week we had two great webinars as part of the HACID series: 1. Collective intelligence in the era of LLMs 2. Scaling Climate Services & Smart Agents for Decision Support If you missed them, stress no more! Watch the recordings on the project webpage:

New webinar announcement! On April the 29th, we will explore how AI can support the growing domain of climate services for adaptation management, and democratise access to detailed climate science information. For more info and registration:

In this webinar, three experts in the field—Jason Burton, Lucie Flek and Nicolas Zöller—will discuss the unique opportunities and challenges this transformation poses for collective intelligence. After that, webinar participants can engage in a final panel discussion session.

Collective intelligence underpins the success of groups, organisations, markets and societies. At the same time, large language models (LLMs) are transforming how information is aggregated, accessed and transmitted online.