Gaurang, you make several important connections between Firestein's point on ignorance and societal problems of climate change and mental health. In each of these problems, I think human behavior poses a significant unknown. How do people react to climate change policy (e.g. support vs. against)? Do they only react strongly to economic incentives, or can the public be shaped to care about the environment as their own, rather than a public good in the tragedy of the commons? Also for mental health, there is likely not be a one-size-fits-all set of indicators for mental health risks or disorders. Perhaps uncertainty can be diminished by more personalized approaches, potentially powered by machine learning and big data.

Daniel, you raise an interesting question. It seems that as time passes and an earthquake still has not occurred in a region, we would expect it to occur sometime sooner than later. However, this assumes that the earthquake occurring is dependent on when the previous earthquake occurred. If the earthquake occurring is a completely independent event, then we would expect uncertainty to remain relatively constant. However, intuitively there seems to be some sort of dependence, which can be modeled by statistics. Perhaps using past earthquake data to fit the number of earthquake events to a distribution (e.g. Poisson distribution for an event with low probability) could help to model the change in uncertainty over time.

I appreciated the real-world examples of the combinations of risk and uncertainty (whether high or low for each). They concretely demonstrated the difference between risk and uncertainty in a way accessible for all readers. Perhaps including a graphic with the example scenarios (e.g. 2x2 table of risk on horizontal axis and uncertainty on vertical axis, similar to a Prisoner’s Dilemma table) would be a quick way to reference the difference between risk and uncertainty. Overall, I thought the essay addressed an important topic, since risk and uncertainty are often used interchangeably in everyday conversation but have different meanings.

Hi Ben, you raise an interesting question about quantifying uncertainty in the Drake equation. It would be nice to use historical data to quantify the uncertainty of each term of the Drake equation; however, there is no historical data of extraterrestrial civilizations. Perhaps another way to measure uncertainty could be to run simulations to determine the distribution of each term in the Drake equation. Each term, such as the fraction of stars that have planets, corresponds to a specific process (e.g. planet formation), which we can simulate by breaking it down into sub-processes (e.g. rocks crashing into each other in space and sticking together to form larger clumps). After building these simulations from fundamental processes, we can measure the distribution (such as mean and standard deviation) of the desired term to better understand its uncertainty.

Aurora, you’ve highlighted an intriguing connection between prediction and human desire for control. I think predicting something you want to happen often gives you an “illusion” of control that helps to assuage any worries about the future. This is an illusion because the future is still uncertain until it occurs, despite your best estimations. In many cases, predicting what you want to occur may even decrease your "control" as you hold a biased view of the future. I wonder whether predictions that are made more according to future hopes than accuracy should still be called “predictions,” or rather “wishes.” It would be an interesting exercise to examine modern prediction systems to see where this hope introduces inaccuracy and bias. For example, if people don’t want to hear a gloomy prediction about climate change, can the predictor frame it in such a way that people will listen and take action? Moreover, could telling people what they want to hear be a self-fulfilling prophesy, as you pointed out?

You make a great point, Julie! Judging whether data is “good” seems to require examining the use case of the model and pre-existing circumstances that may influence decision making, both by the model and/or by humans interpreting the model. Model accuracy is definitely an important goal to optimize for, but we may also want our model to optimize for some metric of fairness, perhaps based on race and gender. Does the model treat individuals of different races or genders equally? Can we say something about the distribution of results being independent of race or gender? This notion of statistical independence may be helpful to define whether the model is fair. Because the model depends on the training data, "good" data can be defined as data that leads to a "fair" model. Perhaps, we can think of “good” meaning that the data is representative of the population that the model will be used on. For example, is the racial and gender breakdown of people in the dataset similar to the overall population? If not, the model may be biased, as Vox media producer Joss Fong illustrates in her video “Are We Automating Racism?”.

Gavin, I’m glad you brought up the question of recycling. Even if recycling plastics is currently not economically or environmentally efficient, the leftover waste generated in the absence of recycling is another important cost to consider. Government investment and subsidies in recycling research could potentially reduce the economic costs and carbon emissions, just as they did for solar, wind, and other renewable energy sources. Additionally, your point that recycling plastics is not economically viable made me think about how policy and economic incentives can motivate companies and people toward certain behavior. For example, charging 5 cents for buying products with plastic bottles and refunding that 5 cents for recycling is meant to encourage consumers to recycle more often. However, many people simply overlook this 5 cents and don’t bother recycling. Another way to encourage companies and individuals to recycle is to implement something similar to a carbon tax. If you create more waste due to not recycling, you have to pay an additional tax per unit of waste. But these incentives all depend on whether it is worth it to promote recycling.