Learning Science30 July 20266 min read

Why Talking to an AI Can Actually Make You Better at Talking to Humans

It sounds like it shouldn't work. Practicing a fundamentally human skill — reading a room, holding your ground under pushback, thinking on your feet — by talking to something that isn't human. The obvious objection is that you should just go talk to more actual people. That advice isn't wrong, exactly. It's just missing what makes practice effective in the first place.

What Deliberate Practice Actually Requires

Psychologist Anders Ericsson's research on expert performance — the work behind most of what's understood today about how skill is actually built — found that raw experience isn't what separates experts from everyone else. Two people can spend the same number of hours doing something and end up at very different skill levels, because time on task isn't the active ingredient. What Ericsson called deliberate practice has four specific components: a clear, specific goal for what's being improved; tasks pitched right at the edge of current ability, hard enough to require real effort, achievable enough that progress is possible; and immediate feedback that lets you correct errors while the attempt is still fresh in memory.

Most real-world conversation has none of these by default. You don't choose the difficulty. You rarely get honest, specific feedback on what actually went wrong. And the goal of a real conversation is usually the conversation itself, not your own improvement.

Why 'Just Talk to More People' Doesn't Meet the Bar

Real conversations are valuable, but they're a poor training environment for a specific reason: the cost of failure is real, so you avoid the exact moves that would teach you the most. You won't try a riskier rhetorical approach on your actual boss. You won't deliberately let an actual first date go quiet just to practice recovering from it. You won't ask your actual interviewer to challenge your weakest answer twice. The stakes that make real conversations matter are the same stakes that stop you from experimenting inside them — which means the moments with the highest learning value are the ones you're least likely to attempt.

There's also the feedback problem. Most people, most of the time, don't tell you clearly what didn't land. They soften it, or they don't mention it at all, or they decide silently that the conversation just didn't go anywhere and move on. You're left guessing what to fix.

Continue reading — free.

Create a free Soro Soke account to read the full article — and get access to spaced repetition, Speak Up, and Small Talk Lab while you're at it.

Already have an account? Sign in

Put this into practice.

Soro Soke gives you spaced repetition and AI-coached practice — Speak Up, Small Talk Lab, Debate Lab, and Doc Lab — all in one place. Free to use.

Get started free