| Summary: | <p>If life emerges on a planet, is it likely to become intelligent? And if intelligent life does emerge, how long will it persist? These two questions motivate this thesis, and are linked by a common methodological challenge associated with <em>observer selection effects</em>. An observer selection effect is a particular type of sampling bias that occurs when a sample is not representative of all out- comes, but rather a subset of outcomes compatible with the existence of the observers. Naive approaches that fail to account for this potential bias could overestimate the probability that intelligent life emerges or underestimate the rate of catastrophes. An overview of these topics is presented in Chapter 1. </p> <p>Using highly simplified models, I explore ways that observation selection effects could affect our understanding of the history of life on Earth. In some instances, these simple models enable us to conclude that observation selection effects are unlikely to be a major factor. For example, Chapter 2 investigates the rate of human extinction from natural sources, and finds that the rate is reassuringly low and unlikely to be heavily biased by observer selection effects. </p> <p>In other instances, the simple models provide a reason to think that observation selection effects could have an overwhelming impact, opening up new hypotheses for future research. For example, Chapter 3 investigates the timing of key evolutionary transitions on Earth, and concludes that exceptionally rare events occurred much earlier than expected due to observation selection effects. Chapter 4 builds on this result, finding that catastrophes, when combined with observation selection effects, can accelerate hard evolutionary transitions that past literature may have overlooked. </p> <p>Finally, in some cases the impact of observer selection bias may be modest but not overwhelming. In Chapter 5, I examine the rate at which catastrophes might destroy the biosphere, and conclude that although observation selection effects may play some role in underestimating the rate, it is unlikely to bias our estimates enough to be meaningful on the scale of human lifetimes. I finish with a short discussion of some general principles suggested by the work, and directions for future research. </p>
|
|---|