This document summarizes Andrew Gelman's article about the usefulness of p-values. Gelman groups p-values into three categories: strongly useful, weakly useful, and misleading. He provides examples from his own work where p-values were strongly useful in determining an election was fairly run, and weakly useful in a redistricting study where reporting the p-value would have been unnecessary. Gelman also discusses a study by Daryl Bem that misleadingly interpreted p-values to support precognition, when more analysis showed the data did not actually support that hypothesis.