The article discusses a model where people can choose between two ways to predict prices, leading to different outcomes. The selection process is like a game where individuals pick the best forecasting method based on performance. Some solutions are not stable when learning is involved. The model shows that one prediction method is vulnerable to temporary changes in information costs.