Let’s talk about Trendslop.

An interesting new HBR article shares research showing that when you ask an LLM for strategy advice, it biases heavily to the fashionable rather than the specific. And no, better prompting doesn’t address the issue.

The LLMs favor

Differentiation over commoditization

Augmentation over automation

Long term over short term thinking

Decentralization over centralization

Incremental innovation over radical innovation

Collaboration over competition

Exploration over exploitation

The authors have many great tips for avoiding convergence to a big undifferentiated pool of guidance, including better context and pushing for the counterintuitive advice. It also argues that if you give a few options, the LLM will hedge, rank, or blend them.

All very interesting.

Here’s what the authors don’t mention that I think is worth discussing. The vast majority of the prebaked “biases” are PROSOCIAL. Longterm thinking, collaboration, exploration, all favor sustainability, stakeholders, etc etc.

This reminds me of other research showing that LLM advice reduces polarization.

Are the AIs pushing us to more healthy, collaborative leadership?

And, are these ideas actually more fashionable? Because everything else I read is exploitation all the way down!