Sign in Subscribe

I Don’t Trust AI More Than a Clear Thinker But It Can Reveal What I Might Miss

Last updated on 

Context:In high-stakes environments like drug development, public health, and regulatory science, inference is everything. But inference is also tricky. Even the clearest human thinker can draw the wrong conclusion if a key variable is missing. And machines, for all their speed, don’t actually know what they’re doing.

1. Human Inference: Powerful, But Vulnerable

  • Humans draw inferences instinctively, often before they can explain them.
  • We excel at contextual reasoning, ethical framing, and adjusting for edge cases.
  • But we’re vulnerable to confirmation bias, incomplete data, and post hoc rationalization.

2. Machine Inference: Broad, But Shallow

  • AI can process millions of variables and detect subtle statistical associations.
  • It doesn’t get tired or defensive. It can update faster and more consistently than any person.
  • But it lacks understanding. It doesn't know which features matter or why a correlation is biologically implausible.

3. The Real Problem: Unquestioned Authority

  • Machines can be wrong, just as humans can.
  • What makes machine error more dangerous is our tendency to overtrust its objectivity.
  • If a human is wrong, we challenge them. If AI is wrong, we might not even notice.

4. My Strategy: Use AI to Surface the Unseen

  • I don’t ask AI to replace judgment. I use it to show me what I might have missed.
  • I use it to scan for patterns, contradictions, or overlooked variables, and then apply reasoning.
  • I hold machines to the same standard I hold myself: What’s the mechanism? Is it biologically sound? Does it generalize?

5. The Future: Not Replacement, But Partnership

  • Inference should be human-led and machine-informed.
  • The best outcomes come when thoughtful people use AI to challenge their blind spots, not abdicate responsibility.

Closing Thought: Inference isn’t just a technical process. It’s an ethical one. We infer meaning. We decide what matters. AI can help us see more clearly, but it cannot choose wisely. That’s still our job.