Shoot First, Ask Questions Later? Building Rational Agents that Explore and Act Like People