In a sense, a special case of RandomReinforcementSchedule and PositiveReinforcement, as noted by Bruce_Shelly? in Game_Design:_SotS?: "Avoid random events that adversely affect a player. These just frustrate the player and add very little of interest."
It falls under his general rule of MaximizeFunMinimizeFrustration.
This should not be taken to mean that negative events should never occur in any random context; CalculatedRisk would be very hard to create without a careful balance of good and bad outcomes. The important distinction is that the player should have enough information available to him to be able to assess the risk of the Bad Thing happening, he is then making a meaningful choice and will not feel frustrated if the Bad Thing happens.
The key difference is whether or not the Bad Thing about to happen or just happened is arbitrary, created by the game system, or whether it is the consequence of some choice. The more arbitrary and random a Bad Outcome is, the more player frustration it will engender.