, , , , ,

Science is highly concerned with data.  You can collect it by studying the amount of light from a binary star system over time, by looking at the changing thickness of a polymer thin film over time, by seeing how many times a piece of paper can be folded before it cannot be folded anymore.  Whatever an experiment is, there will be data collected from it

And yet, there, in the midst of your data, there is a human [1].

“But wait!” you say.  “Stars are really far away and have in no way been touched by humans.  This data was collected by a computer.  In the case of computer simulations, it was GENERATED by a computer.  How can you possibly be saying that there is a human in my data?”

I can say that because a human built that computer and wrote that simulation.  I can say that because a human built your telescope.  I can say that because several humans wrote your textbooks or academic journal papers.  I can say that because there is a human writing down the numbers or observations.  Humans are there, however distantly, along every step of your experimental design, execution, and analysis.

Because a human is involved along every piece of data collected, there must be an allowance for human error and bias [2].  This is the most basic source of error bars in graphs.  Because we are human and make mistakes, we must reflect that source of error in our data.

Those of you who do their best to be objective, accurate, and precise in science will find this stubborn and ineradicable source of error frustrating.  Have some compassion for yourself and do the work to the best of your ability.  Humans are your primary audience, so you’re in good company.


[1] For simplicity, assume that the experiments I speak of do not involve clinical trials in humans.  There are several splendid double-blind controlled trials out there, but those contain humans automatically.  People are not particles.

[2] Sometimes the error is along the lines of, “Derp, I accidentally hit the table where my experiment is running.”  Sometimes the error is in the way the telescope interprets light coming into it.  Sometimes it is the way a computer program is written.  Sometimes the person looking at the data does not want to see it LALALA I CAN’T HEAR YOU.