The Observational Issue

In a previous article we spoke about the problem of having exceptional individuals as your sample population or as your design team - basically observer's bias skewing the data. But in this one, we address something much more delicate - the issue of the existence of the observer.

An interesting, and a lot more relatable example of the observer's existence skewing sample data can be observed in linguistics. Say you walked into a room of friends chatting casually, and then told them that you were going to start recording their conversation - even assuring them that the recording won't be made public will inevitable bring about a change in their speaking manner - maybe fewer idioms, slang or curse words. This is actually a pretty huge issue in linguistics, where studies depend on the capability to observe the speech of native speakers  - who have their own unique talking style, idiosyncracies, idioms and features that make their interpretation of the language, beatifull. The existence of the observer eliminates that entirely.

A more physics oriented example would be the (in)famous double slit experiment.  The double slit experiment famous involves that electron detecting sheet at the back, but why is it required? The issue that arose was that, placing a detector in front of the slit somehow skewed the data - it seemed like the electrons were actually passing through only one of the slits. Placing the screen helped track the electrons by removing the detector entirely.

Movies also seem to have caught on to this phenomenon - accidental superheroes and supervillains are created only when the brilliant scientist is not looking. Programmers run their code with their eyes closed after debugging it for days on end. The most amateur athletes can pull of shots they can only dream off only when their eyes are closed (now this can actually be attributed to some science - there aren't any distractions that normally affect the aim of the athlete. This technique is pretty  untrustworthy though, stuff might go in the exact opposite direction more often than not).

But into a professional setting agian, surveys are majorly affected by this. Knowing who the person taking the survey is, can always skew the responses as most surveys are never fully anonymous. Words said by the survey-taker can reside in your memory and skew answers based on the wording of the questions themselves. The human mind is capable of skewing its viewpoint based on even the smallest things - these can include trigger words, using the word down, will divert some surveyees towards the last options, surveyors wearing a particular colour clothing can skew the surveyee towards or away from it. Poor grammar in a question, can disinterest the surveyee, or even appearing too amiable, or unprofessional, or sometimes appearing too professional for light hearted surveys who's answers are supposed to involve lesser thinking.

The observer's bias and existence are actually two huge problems in understanding populations - for science, research and product building - an unfortunate but almost unavoidable risk as part of the trade, but one that can be reduced as much as possible - through a miz of anonymity, and the prevention of idea-skewing prompts.