Contrarian Incentives and Costly Social Learning
We study social learning when agents choose costly information and prefer less popular actions. Popularity changes decision cutoffs and can restore the value of information. Under diminishing, nonsummable popularity updates, we characterize complete belief learning by positive attainable net information value at every interior belief, allowing vanishing entry fees and experiments approaching no information. With Gaussian signals and power precision costs, learning is complete exactly when contrarian incentives can align every belief and costs vanish faster than the square root of precision. Under weaker incentives, optimal purchases can have a positive precision floor and a finite expected count even without an entry fee. Rapid popularity updating can also stop learning despite profitable alignment. A positive fixed fee uniformly bounds purchases; a binary illustration characterizes terminal errors and separates belief accuracy from the frequency of correct actions.