Maybe the next time the survey should not allow probabilities exactly 0 or 1 (rather than saying they’ll be interpreted as 0+epsilon and 1-epsilon), and give the option to express probabilities as log-odds if they’re extreme. (Anyway, I didn’t give any probability lower than 0.1% or higher than 99.9% in my answer.)
Maybe the next time the survey should not allow probabilities exactly 0 or 1 (rather than saying they’ll be interpreted as 0+epsilon and 1-epsilon), and give the option to express probabilities as log-odds if they’re extreme. (Anyway, I didn’t give any probability lower than 0.1% or higher than 99.9% in my answer.)
… why not log odds all the time?