I think it genuinely doesn’t make sense to say that α reflects our prior expectation of α; the acolyte is correct. What α reflects is our prior on h; that regularization term corresponds exactly to a prior that makes h (multivariate) normally distributed with mean zero and covariance 1/α times the identity (i.e., components independent and each component having variance 1/α).
I think it genuinely doesn’t make sense to say that α reflects our prior expectation of α; the acolyte is correct. What α reflects is our prior on h; that regularization term corresponds exactly to a prior that makes h (multivariate) normally distributed with mean zero and covariance 1/α times the identity (i.e., components independent and each component having variance 1/α).