Optimal search with unreliable and dangerous cues

Carbon dioxide is a broad signal of molecular decay, and it is almost universally attractive among insects in search of hosts, wildfires, flowers, decaying matter, communal nests, predators, and fruit. CO_2 is also, however, toxic at naturally occurring high concentrations. It is not clear how insects balance the information provided by this broad and dangerous signal with the information provided by odors that are more unique to their respective niches. This particular challenge is an example of a common dilemma that all animals face. To address this question, we studied how fruit flies balance the value of information provided by CO_2 and ethanol, both important odors produced during fermentation, their primary food source. We found that flies exhibit similar attractive responses towards CO_2 as they do towards ethanol, however, they invest twice as much time in searching near sources of ethanol. To understand what these differences in search times might mean in terms of their ecology, we simulated different virtual ecosystems and found that their strategy is optimized for scenarios where CO_2 and ethanol correspond to 30% and 70% chances of finding food, respectively. Our simulations extend beyond this particular case study by providing a plausible explanation for why experimental observations so often do not agree with predictions of optimal foraging theory and the marginal value theorem. Curiously, our result that flies find CO_2 attractive runs contrary to the majority of the scientific literature, which has suggested that flies find CO_2 aversive. In our experiments, we did find that flies occasionally do find CO_2 aversive, but only during times of low activity. This could be an adaptation to reduce the chances of falling prey to parasites or lethal concentrations of CO_2