We present an experimental design where uncertainty is generated from the advice of experts with conflicts of interest. In this experiment clients are faced with a variant of a multi-armed bandit problem with a random end-time. On the known arm (the “task screen”), clients can earn a certain payment per completion of a decoding task. However, clients may also opt for the unknown arm where they earn an uncertain amount if they end the experiment on this “expert screen”. The amount is uncertain to the clients because the value is being communicated through an “expert” with conflicted incentives. A control session provides for direct transmission of the value to the clients. Our results show that ambiguity aversion is alive-and-well in this environment. Also, when we vary the wage rate on the known arm we find that higher opportunity cost clients are less likely to heed the advice of conflicted experts.
