iiinova — where your digital self keeps evolving

Carlo Reverberi

@carloreve.bsky.social

Decisions, AI, brain and all the rest

Carlo Reverberi @carloreve.bsky.social · 27/05/2026

The future of human-AI collaboration isn't just about building smarter AI. It’s about understanding and improving how humans navigate the advice they receive!

000

Carlo Reverberi @carloreve.bsky.social · 27/05/2026

Optimal integration matters! Our data shows an accuracy gap of up to 12% between the top and bottom tertiles, highlighting major individual differences in performance.

100

Carlo Reverberi @carloreve.bsky.social · 27/05/2026

Two possible mechanisms behind this: Ego bias: people tend to overweight their own judgments relative to those of others, even when the other source is better. Global-to-local integration: people struggle to translate that global knowledge on advisors into the best decision on a specific trial.

100

Carlo Reverberi @carloreve.bsky.social · 27/05/2026

The less good news: They were still far from optimal. When advisors were better than themselves, participants often left substantial performance gains on the table. In other words: good advice was available, but not fully exploited.

100

Carlo Reverberi @carloreve.bsky.social · 27/05/2026

The good news: People were often adaptive. They could improve their performance by using advice and were especially good at rejecting poor advisors. So humans are not simply “blindly following AI”.

100

Carlo Reverberi @carloreve.bsky.social · 27/05/2026

Participants made perceptual decisions while repeatedly interacting with the same advisors over time. These advisors varied in ▶️ accuracy ▶️ confidence ▶️ metacognitive sensitivity The environment was transparent: participants received feedback and had access to global information about each advisor

100

Carlo Reverberi @carloreve.bsky.social · 27/05/2026

In our new paper, we asked a simple question: When people receive advice from AI-like agents, do they integrate it optimally? The answer is: partly yes, but not enough. www.nature.com/articles/s44...

nature.com

Adaptive yet suboptimal integration of advice in decision-making - Communications Psychology

Advice improves decisions, yet its use is suboptimal. Combining experiments and Bayesian modeling, this study identifies egocentric discounting and a failure to apply global traits to trial level choi...

110

Carlo Reverberi @carloreve.bsky.social · 27/05/2026

💣 PAPER OUT! 💣 We live in a world full of (AI) advice. But having access to good advice does not automatically mean making better decisions. In many contexts, the final call is still human! #HumanAICollaboration #DecisionMaking #Advice #AdviceIntegration #MetaCognition

120

Carlo Reverberi @carloreve.bsky.social · 12/09/2025

Great work @marcociapparelli.bsky.social ! Well deserved publication! 😃 And with CerCor nice editing now reads like a charm!

010

Carlo Reverberi @carloreve.bsky.social · 02/05/2025

@giosuebaggio.bsky.social this is perfect for you!

000

Carlo Reverberi @carloreve.bsky.social · 02/05/2025

Brilliant preprint by @marcociapparelli.bsky.social on compositionality in the semantic network. He used model-driven RSA to identify brain regions representing building blocks vs. compound meanings in concepts like cat-fish or cactus-beetle #compositionality #rsa Details below! 👇

120