Abstract
When people anticipate possible future outcomes, they often prefer to know their fate in advance. More for-
mally, gambling-based experiments into what is called ’observing’, or ’information-seeking’, indicate that delayed
stochastic rewards become more attractive when the uncertainty about the outcome is resolved in advance by a
predictive cue. Recent monkey experiments showed that neurons in the lateral habenula responded differently to
the same predictive cues depending on how expected these cues were (Bromberg-Martin and Hikosaka, 2011).
No existing model offers a satisfactory account for these data. This includes ideas related to Shannon information,
as targets conveying less information can be more attractive (Roper and Zentall, 1999). Further, the preference
appears to depend on the delay length between predictive cues and rewards (Spetch et al., 1990), a factor of
little informational consequence. The notion that subjects are differentially engaged whilst waiting for the outcome
according to the cues (Beierholm and Dayan, 2010), foundered on the recordings in lateral habenula (Bromberg-
Martin and Hikosaka, 2011). Here, following Loewenstein (1987), we considered the delay to a future reward as
itself being appetitive (utility of anticipation or savouring). In addition, inspired by animals’ excitement following the
predictive cues (Spetch et al., 1990), we hypothesize that the attractive force of the anticipation is boosted by the
reward prediction error (RPE) occasioned by the predictive cue. Our model accounts for a wide range of existing
neuronal and behavioral data, including those in (Bromberg-Martin and Hikosaka, 2011), without appealing to
an explicit value for information. We also confirmed the model’s central predictions in our new human empiri-
cal studies, in which subjects strikingly preferred delayed consumption of revealed stochastic rewards. Further
studies also show that subjects seem to be willing to sacrifice rewards to purchase revelation. We suggest that
RPE-boosted anticipation plays a crucial role in driving risk-seeking behaviors