What is Science, Anyway?

Add replication to the quality petal!

During a conversation with a sister, I discovered we viewed the definition of Science differently. I see Science as a careful, prescribed process of discovery, while she saw it as the results themselves. It seems both definitions are used.

Oxford: * the intellectual and practical activity encompassing the systematic study of the structure and behaviour (British spelling, of course) of the physical and natural world through observation and experiment

*a systematically organized body of knowledge on a particular subject. “the science of criminology” 

I doubt the basic qualitative requirements for Scientific studies will ever change, although new research tools may come with their own specifications for rigorous use. In terms of evolution of knowledge, most is in terms of learning something new about our world, but some results contradict long held principles. An example of the latter kind of evolution is related to a recent blog of mine about muons shaking up the world of nuclear physics— the area of Science I believe most filled with unanswered questions. Since the unexpected level of magnetism of these muons contradicts theory whose roots go back to Einstein this could be a revolution rather than evolution. Scientists await further independent verification, but I’m betting many are looking for corrections in the understanding of nuclear physics.

This is good place to warn the reader about bad Science. An unfortunate circumstance related to media’s love of what’s sensational results in undue publicity to unreplicated studies contradicting accepted norms or fun results like one glass of red wine is worth an hour of exercise. Beware of “Studies show …” without reference to a reliable source as preached in a previous blog.

Mathematics is the tool of Science—sometimes called the Queen of Science. Some is heavy duty advanced mathematics as in quantum mechanics, but most common is elementary statistics. Replication increases the probability of the accuracy of preliminary results. Not only is the media guilty of jumping the gun on a single study, unscrupulous or unschooled researchers seek random correlations by studying many attributes of a sizable sample of say, people. Probability predicts that two or more attributes will coincidently appear correlated. (It is a fact that in a group of 35 people or more, the odds are that two or more have the same birthday. To get the probability that it doesn’t happen one multiplies 35 or more numbers less than one. Then it’s just arithmetic that the product will be small.) Similarly, some correlation of attributes is expected in a fishing expedition with lots of fish, making it illogical to make any conclusion about the discovery of the apparently correlated without replication. The mistake of concluding correlated qualities involve cause and effect is even a more common error in popular communication of scientific studies.

The advertisements of many health supplements use language crafted to sound scientific. Sometimes unjustified claims are scientifically studied and debunked for the sake of the public. This naturally adds to the sense that Science is in flux.

Weapons of Math Destruction?

mathweaponAs a retired mathematician, how could I resist this book title? When I read that economic inequality was furthered by use of mathematics, I was even more intrigued. I’m also a retired political activist. In fact, I’d like to retire and wake up on November 9th.

My first thought was that like any tool, mathematics could be used for harm or good. One can use a hammer to work for Habitat for Humanity or do someone serious harm. I was being a prig in taking the clever title too literally.

Okay, so how is math being used to keep the poor, well poor? According to the author, Cathy O’Neil, algorithms and big data, intentionally or not, target the poor, reinforce racism, and amplify inequality. In a court of law or in a general education course on critical thinking, guilt by association is decried as irrational. Yet, that’s exactly what happens when one’s zip code is used to determine loan eligibility or  a loan’s interest rate along with home and car insurance rates.

Employers use credit scores to measure responsibility, but this equates dependability with higher income. If your credit record is due to unemployment, your unemployment keeps you unemployed.  In order to prevent unfair bias in sentencing, recidivism models were adopted by some states. However, the model includes criminal records of friends and family.

Personality tests have been devised for employment purposes. The math presumably only figures in the final score of acceptable answers. It’s not clear math has been used to verify a correlation between passing and performance on the job. The tests have been accused of measuring averageness, thereby denying outliers whose creativity could be an asset. This simply adds up (pun intended) to unfairness across the board rather than a bias against the poor.

McNeil calls these tests WMDs and labels them opaque and unfair. There is no explanation of what went wrong when these measures are used to deny life benefits.

Not mathematics, but the use of statistics is responsible for this judgment by association. More precisely, if you have a zip code, credit score, or other identifier that a software program places you in a demographic less likely to succeed,  it becomes a self-fulfilling prophecy.

I read a very interesting review of the book that made me return to my thought about a tool’s capacity for good or harm. The reviewer had been poor and later lived in a neighborhood with a favorable zip code. While the reviewer suffered from the algorithms mentioned in the book, he also talked about some of the positive aspects of the tools. He appreciated more police deployment in his high crime zip code and a mechanical measure of tardiness, which prevented bosses from forgiving their pals.

If nothing else, Ms. O’Neil has heightened awareness of built-in biases in programs where we expect objectivity.