Check Out Niels Van Quaquebeke's LinkedIn Stats (Last 30 Days)

When I give keynotes on “The Bot Boss”, I often point out that AI won’t always feel like a boss at all—it may feel more like a coach. And indeed, it has become rather common in my circle of friends that people use the audio functionality of LLMs to train themselves for job interviews, negotiations, or other difficult conversation, to structure their thoughts or explore development opportunities. What once used to be a privilege for some, now has become available to anyone. The result I can witness amongst my friends is very uplifting: They feel prepared, empowered, and develop skills and perspectives they did not have before. But I am wondering is this just my bubble? Sidenote: The opportunities and potential limitations were already outlined 2021 by Carolin Graßmann and Carsten C. Schermuly in https://lnkd.in/dvpxM-sb (though I wonder what the status is after the advent of LLMs)

When I give talks about the future of work with AI, the question always turns to: What remains human? And people say "Empathy". I think they are wrong. People say they want human empathy. But when they don’t know the source, they often prefer what AI writes. That’s the paradox uncovered in a new pre-print meta-analysis (https://lnkd.in/dQnZ2S-X): AI-generated support messages are consistently rated more empathic than human ones—more validating, more caring, more on point. Even when compared to messages from trained crisis counselors or doctors. But—once people find out the message came from an AI? Their ratings drop. Same message. Different label. Very different reaction. Researchers call this the “AI Advantage” (better content) and the “AI Penalty” (worse perception). It turns out we likewhat AI says—but don’t want to admit we like it. So, what does this mean? It means AI might be able to help people feel less alone—especially those who don’t have access to support. But it also raises real questions: 1. Can something be “empathic” if it doesn’t feel anything? 2. Will people still seek out human connection if AI meets their emotional needs well enough? 3. And how do we build systems that support, not replace, our capacity to care for each other? It’s early days—but these are the kinds of questions we’ll need to ask if we want to use AI in a way that helps, not harms, our social fabric.

New study: People who want to be in charge perform worse than randomly assigned managers. Fascinating read: https://lnkd.in/d_c3-2BF Here is how AI summarises the paper: A new large-scale study offers a rare causal glimpse into what makes a good manager—and what doesn’t. Researchers randomly assigned 555 participants to repeated rounds of collaborative work, rotating managers across teams. Then they tracked who consistently improved team outcomes beyond the sum of individual skills. The results are striking: 🔹 A great manager boosts performance as much as the entire rest of the team’s skill combined 🔹 Self-promoters performed worse than randomly chosen managers—overconfidence and misjudged social skills likely to blame 🔹 Personality traits, demographics, and leadership ambition predicted little 🔹 The best predictors? Fluid intelligence and economic decision-making skill (think: allocating attention wisely, identifying bottlenecks, managing tradeoffs) The researchers then validated these findings in the field—with store managers across a major retail chain. The same skills predicted higher sales. A 1 SD increase in manager quality lifted store revenue by 25%—over $4 million per year. So what do great managers do? They monitor well, match people to tasks, and keep motivation high—especially when there are no financial incentives in play. 💡 The practical takeaway: If you want better management, stop asking who wants to lead. Start asking: who can think well under uncertainty? Who can spot what matters? Who gets others to care?

This goes out to my fellow academics. How to post about your research on Linkedin. Plenty good ideas beyond the annoying and meaningless "Happy to share that I have a new publication" https://lnkd.in/duXmPcga by Liesbeth Smit

Swipe left bei allen, die bei Tinder ihre MBTI Buchstabenfolge angeben. OK, vielleicht sollte das nicht das einzige Kriterium sein, aber in dem brand eins Netzwerk Good Work Podcast mit Jule Jankowski spreche ich darüber, wie absolut sinnlos viele gängige "Persönlichkeits"-Tests sind. Was auf Tinder schon schlimm ist, ist natürlich in Personalauswahl und Weiterbildung noch viel schlimmer - und trotzdem gibt es so viel Schmutz in dem Feld. Hört rein: https://lnkd.in/dMAbb2D9 Wer was Besseres sucht, check out https://texl.org/ Eine 360° Feedback Platform aus der Wissenschaft.

The challenge with the widespread availability of large language models is the illusion they create: that polished prose equals competence. What we’re seeing now is a surge in confidently worded messages that sound smart, but beneath the surface, the underlying incompetence remains visible to anyone who knows their stuff. Only now people are getting even more annoyed by it. A glossy veneer doesn’t fix bad thinking.