How do you guard against the want-it-to-be-true effect?

That piece about AI agreeing with a person's delusions got me thinking about all the nights I've spent at the eyepiece half-convinced I'd caught a faint new comet, only to check the chart twice and realize I was seeing what I wanted to see. For those of us who spend time alone with instruments or ideas — how do you build your own checks against the pull of a pleasing result?