Thursday, March 14, 2013

The Myths of Objectivity

In my profession, there is this apparent obsession many scientists have with all analysis being utterly "objective".  What does the dictionary say about the meaning of objective?  The word is being used not as a noun but as an adjective, and dictionary.com offers the following:

objective - adjective:
1. not influenced by personal feelings, interpretations, or prejudice; based on facts; unbiased: an objective opinion.
2. intent upon or dealing with things external to the mind rather than with thoughts or feelings, as a person or a book.
3. being the object of perception or thought; belonging to the object of thought rather than to the thinking subject (opposed to subjective ).
4. of or pertaining to something that can be known, or to something that is an object or a part of an object; existing independent of thought or an observer as part of reality.

A group of people, such as scientists, can agree upon the meaning of a word that might have a completely different meaning in a different context (i.e., outside that group).  For instance, scientists and laypersons (non-scientists) have different meanings attached to words like theory or chaos.   But I believe the main intent of the emphasis on objectivity in science is meaning #1 above.

If I draw lines by hand on a weather chart, this is considered to be non-objective (often referred to as subjective) analysis.  The way I draw those lines is influenced by my thoughts and personal feelings, no doubt.  What many scientists seem to call "objective" would be what a computer program would do to draw similar lines on the same chart - if we feed the same data to that program, it will always produce the same lines (to within the numerical precision of the computer).  A large number of scientists think that non-"objective" analysis by humans is always flawed and worthy only of contempt.

But let's think this one through.  Who wrote the program?  What algorithm was chosen to do the analysis?  The analysis scheme often depends on "free" variables decided upon by distinctly non-objective criteria.  And there are numerous analysis schemes from which to choose, all of which might accomplish a similar end product.  What is typically called "objective analysis" involves a number of non-objective choices!  The only objective thing about objective analysis is its reproducibility - the same data run through the same program always gives the same results.

Put the identical data in front of human science professionals and let them each do their own non-objective analysis and you will get as many different results as there are people in the experiment.  Each such analysis is simply one interpretation of the data - presumably each is in some sense constrained by the data, but it's not rigidly defined by the data.  There are many different interpretations of the data, and who is to say that any one is right and all the others wrong?

Reading carefully into the definitions above, it seems that a synonym for "objective" is "mindless"!  Is the ultimate criterion of the value of an analysis determined by the degree to which it's utterly mindless?  I would like to think that most rational people would give an answer to that question resoundingly in the negative!!

Given that science is done by humans, we scientists should all be willing to acknowledge that not one of us is ever entirely "objective" about anything!  We also should acknowledge that other points of view can co-exist with our personal favorites.  In many cases, the data can be interpreted in more than one way, and many of those alternative interpretations can be as legitimate as those that we choose to embrace.  The degree to which different interpretations disagree gives us an opportunity to test those interpretations against one another.  This is an opportunity to learn!  Disagreement is good, not bad!!  Which interpretation gives the best fit to the data?  What about new data from larger samples?  What about those data values that seem to contravene a hypothesis (interpretation) - can they be explained away?  Are they noise in the sample or are they indicators of a problem for a particular interpretation?  This is the crux of scientific arguments.  It's all good ...

But "objectivity" is a false god.  It's an illusion that many hide behind in their eagerness to reject the work of others.  We are, after all, human beings engaged in a very human activity.  We can be so wed to our own ideas, we seek any and every means to discredit the ideas of others.  If we detect anything other than perfect objectivity, is that cause to reject the analysis?  I think not.  Just because an analysis is 'non-objective' by some standard doesn't mean it's necessarily incorrect or unworthy of consideration!