A network for students interested in evidence-based health care

# Understanding Uncertainty: a resource for scientists, journalists, and the public

Posted on May 12, 2015 by Katherine Stagg

Learning Resources

Scientific research is not political, but the implications of it can be.

Making Sense of Uncertainty is a fascinating paper that takes the reader on a journey through what uncertainty in research means and how it can be understood in order to make decisions, or used to misrepresent or devalue scientific arguments, whether intentionally or through misunderstanding.

Created by Sense About Science, an organisation dedicated to improving public understanding about scientific issues, and featuring the personal views of over 20 academics from a variety of fields, Making Sense of Uncertainty is a thorough introduction to the world of scientific unknowns that could be enjoyed by the layperson and professional alike. It takes about 30 minutes to an hour to get through but is well broken up into different chapters, with practical examples frequently given which really helps the topics to be understood in relation to real-life research.

Chapters cover what uncertainty within science is, whether or not it matters, how we can use results despite uncertainties and how scientific models are affected.

## The (Un)Certainty of Statistics

Many people dread the use of statistics but many others are adept at using statistics to convey the message that they want. A lack of confidence with statistics is a key reason why many shy away from looking at scientific evidence for themselves, and it also allows those who do understand statistics to potentially manipulate them in misleading ways. Making Sense of Uncertainty doesn’t give an in depth explanation of different statistical methods and their use (this has already been done elsewhere by Sense About Science, and the reader is directed there), but it does give both practical and theoretical advice about how difficulty with statistics can be overcome. For example, separating short-term and long-term effects and improving experimental design.

An interesting idea discussed is that in order to quantify your unknowns, you must actually already know a fair amount about a topic. The example given is that if you know there is a 20% chance of rain then you must know about the different variables that make rain less likely, and how they interact so that one in five times there will be rain. Indeed, knowing what you don’t know is very useful, because it allows experiments to be designed around those constraints. Arguably it is the unknown unknowns that are more dangerous, because you are not aware of their impact on your results and so cannot account for them.

## Manipulation of Models

Scientific models are not something that the average healthcare student gets much, if any, exposure to but they are hugely powerful and used in many different aspects of medicine, from the molecular modelling of cellular processes to the prediction of disease spread on a global scale. I was really interested by the chapter dedicated to explaining how scientific models work and how they deal with unknown factors. Models have the advantage over conventional research that they are able to handle very large amounts of data simultaneously and their inputs can be adjusted to see how the result changes much more easily than in practical experiments.

The problems with scientific models are also discussed, namely their use of assumptions and the difficulty in keeping track of how a model has been changed over time. This means that there is not only uncertainty in the results that a model produces, sometimes there is uncertainty in how it got that result in the first place. That kind of uncertainty is not helpful, since it prevents the recreation of results by independent researchers and should be avoided by clear reporting of how a model has been created and fine-tuned over time.

The question is not do we know everything, but do we know enough?