Power in Prediction – Part 1: A look through history

Read more

Read full article

Predicting the future: an ancient art

The desire to predict the future is by no means a new concept. Our progress, however, has been largely accelerated by our mathematical and computational power advances, meaning that today, analytics and predictive analytics are embedded into our everyday lives. The success with which these technologies have been utilized, has seen the healthcare industry adopt similar approaches to improve patient care, and optimize diagnostic decision-making. Indeed, in an age where we generate more data than ever before, the potential of predictive capabilities is seemingly limitless.

While we now think of predictive analytics as a complex, data-driven practice, it came from humble beginnings, and did not involve mathematics at all. To understand where we are now, it is useful to look back.  Ordinary, everyday objects such as books, birds, and even cheese were once used as tools to predict the future.1 Back in Mesopotamia and classical Greece, internal organs from sacrificed animals were harvested and inspected for omens. The liver was the most popular tool; while a healthy organ promised a positive future for its beholder, a missing lobe marked the arrival of almost certain doom.1 Similarly, in ancient China, animal bones (commonly cattle shoulder blades) were the predictive tool of choice. The Chinese cracked these ‘oracle’ bones with hot metal rods to read the results and make their predictions.1,2

The Romans preferred to use living animals, specifically birds, to make their predictions. This practice, known as ornithomancy (divination using birds), predicted future events through the study of bird numbers, flight and cries.1 Another popular method throughout Europe; bibliomancy, included the practice of asking a question, opening a book on a random page and using the first visible passage to interpret an answer. The most commonly used books for this practice included the Bible, Koran, Book of Psalms and the works of Virgil, the Roman poet.1

Perhaps stranger still, in the Middle Ages, young women would write the names of their potential future suitors on pieces of cheese and the first piece to mould would reveal their destined partner.1 Considered together, prediction methods have proven to be both varied and highly creative in their methodology. It wasn’t until the mid-17th century and the birth of probability mathematics, when there was a shift away from these traditional approaches and a move towards mathematically-derived predictions instead.3

The birth of probability mathematics

It was 1654 that mathematics was first used to strengthen our predictive capabilities.3 Blaise Pascal and Pierre de Fermat, via a series of letters to each other, debated how to divide a winning stake between two gamblers whose game of heads or tails was interrupted before its finish.2 To solve the problem, the two mathematicians considered all possible outcomes for the unfinished game and in doing so, reasoned which player would be the winner, marking the birth of modern probability theory as we know it.2

Since the efforts of our predecessors, and the new-age thinking of Pascal and Fermat, we have come a long way.2 The subsequent acceptance of significant benchmarks for testing hypotheses further built on these early foundations and strengthened the validity of mathematically-derived predictions.2 Indeed, in 1960 Eugene Wigner commented on the “unreasonable effectiveness” of mathematics in science and noted the “miracle of appropriateness of the language of mathematics” as a “wonderful” gift.4 Seemingly Wigner’s words still ring true today. Predictive mathematics have allowed us to make several predictions of note, and most importantly, to make them with remarkable accuracy.2 We have successfully predicted the existence of Neptune, black holes and even the future location of comets; with such accuracy we were able to land a space probe on the comet.2

Our progress with mathematical analysis has led to an explosion in its application, to the extent that today, analytics, often referred to as Artificial Intelligence (in its ‘narrow’, ‘general’ or ‘super’ forms), or Machine Learning, are commonplace across almost all aspects of our daily life.2,5 They are used to predict changes in emergency hospital care needs, the stock market, by insurance brokers to assess risk, to predict the weather, to determine which inmates of a prison system are eligible for bail, to find missing people at sea, to tailor advertising, and in sports analytics and player performance.2,6,7

This vast applicability of analytics and predictive analytics has created huge waves in several areas. For example, Billy Beane, general manager of the Oakland Athletics baseball team made the headlines when he implemented an entirely statistics-led approach to building a competitive team with only a very small budget.2,8 In doing so, he established ‘moneyball’; the statistics-driven theory that small-value teams can compete effectively by buying players with assets that are undervalued by other teams and selling those players with overvalued assets.9 These principles have also now been implemented across other industries including music and law enforcement.2,8 Analytics have even been used to make predictions about our future actions before we know about them ourselves. In 2013, Facebook engineer Lars Backstrom and Jon Kleinberg of Cornell University teamed up to publish a paper reporting the factors contributing to the success of a long-term relationship; citing mutual friends and photos together as key contributors.10 A little worryingly, the duo were able to use these factors to predict the breakdown of online relationships, with 60% accuracy.10 Almost certainly featuring in the marketing program of a dating app near you!

Predictive analytics may also be used to make court bail decisions. As discussed in Malcolm Gladwell’s book, ‘Talking to Strangers’, when deciding bail character assessment , human judges are often flawed.6 In one study in New York, a computer algorithm chose to grant bail to 400,000 defendants from a database of 554,639 cases, that were 25% less likely to commit a crime than the 400,000 defendants who had been selected by the judges.6 The computer also flagged a high risk group of 1% of the defendants and predicted that, should this group be granted bail, more than half would commit another crime. The judges were less successful at detecting such high-risk individuals, and in their selection, chose to release nearly half of them!6 In this instance, objective predictive analytics could be a useful support tool for objective decision-making in a field notoriously clouded by the potential for subjectivity.

As technology has evolved, it has naturally become more entrenched in our private lives and in doing so, our personal information has become more available than ever before.10 As a result, the potential for predictive analytics continues to expand, and we are likely still a long way, from reaching the limits of this technology.

Want to learn more? See our article Power in Prediction – Part 2: a need for data and analytics in the future. Explore respiratory_care v2.0 and have a read of our whitepapers on small steps towards big dataredefining the value of healthcaretechnology and healthcare collaboration and a need for accessibility.

References

  1. Lovejoy B. 10 historical divination methods for predicting the future. Available at: https://www.mentalfloss.com/article/585258/historical-divination-methods-predict-future. [Accessed January 2020].
  2. ‘Prediction by the numbers’ transcript. A Netflix documentary. 2018. Available at: https://www.pbs.org/wgbh/nova/video/prediction-by-the-numbers. [Accessed January 2020].
  3. APS News. July 1654: Pascal’s letters to Fermat on the ‘Problem of points’. Available at: https://www.aps.org/publications/apsnews/200907/physicshistory.cfm. [Accessed January 2020].
  4. Mitchener WG. The nature of mathematics. 1996. Available at: https://services.math.duke.edu/undergraduate/Handbook96_97/node5.html. [Accessed January 2020].
  5. ANI: Artificial Narrow Intelligence. Available at: https://www.astutesolutions.com/ani/artificial-narrow-intelligence. [Accessed February 2020].
  6. Gladwell, M. Talking to Strangers. London; Penguin Books Ltd. 2019.
  7. Buch VH, et al. British Journal of General Practice. 2018; 68(668): 143–144.
  8. Imagine Sports. The life and career of revolutionary baseball GM Billy Beane. Available at: https://imaginesports.com/news/revolutionary-baseball-gm-billy-beane. [Accessed December 2019].
  9. Grier K, Cowen T. The economics of Moneyball. 2011. Available at: https://grantland.com/features/the-economics-moneyball/. [Accessed January 2020].
  10. Dickson EJ. Can Alexa and Facebook predict the end of your relationship? 2019. Available at: https://www.vox.com/the-goods/2019/1/2/18159111/amazon-facebook-big-data-breakup-prediction. [Accessed January 2020].

March 2020 RESP-42107

Related articles

Stay informed

Stay informed with our monthly updates which contain the latest information on the future of connected respiratory and healthcare innovation.