Wharton Professor Kartik Hosanger dives into AI, and what both performing and underperforming can have on the trust people place in recommendation engines and other applications.