Subscribe to DSC Newsletter

Data is not facts - the impossibility of being unbiased

Best intentions

We talk a lot about making decisions based on data but we need to be careful about how hard and fast those decisions are. Our decisions are only as good as our data and our analysis.

Neither can be perfect. Data is always a sample of the full scope of reality and analytics is always an interpretation of that sample.

We need to be cognizant of the differences between Opinions, Facts and Conclusions. And, just as important, we need to recognize the relationship between our judgement and our ego: all disagreements are personal to some degree. 

Opinions vs Facts vs Conclusions

Opinions are strictly preferences, no reference to supporting data: "Blue is my favorite color", "Bananas taste good".  An opinion is fundamentally about an answer to the question "What do you like or dislike?". 

Facts are incontrovertible characteristics of the world. But the basic bias of our perception and imprecision of our measurement systems mean that the data we collect is very rarely a fact. Instead it is an approximation of an aspect of a fact. 

Conclusions are based on our evaluation of the available data. Note, "data" not "facts". They are a reasonable hypothesis derived from our judgement of imperfectly precise data using our past experience, knowledge, and culture. 

If you think that conclusions can ever truly be unbiased, think again. 

Justification vs Analysis

Humans being what they are, we make all our decisions based on feelings and then find data to support it after the fact. 

Think about how often you change your mind. Even if we are presented with data that contradicts what our belief or decision is, our tendency is to discount the data before we change our minds.

And that's actually not a irrational approach. Again, data is not facts. We have to use our judgement both to draw conclusions from data as well as to judge whether that data is trustworthy or not in the first place. 

Just because I read on the internet somewhere that Donald Trump eats babies doesn't mean that I am going to accept that as being true. In my judgement it is more likely that particular data point is incorrect than it is that we have presidential candidate who is a cannibal.

How do *you* know we landed on the moon?

Most of what we trust as facts are actually conclusions based on an acceptance of the validity of a particular set of analyses of a particular set of data. 

We don't have time to go through and recreate every experiment to verify its conclusions. We just compare its findings with what we already already judged to be true. But our judgement is based on a series of previous conclusions made from that same judgement and thus it becomes self-reinforcing. 

We rarely change our minds when presented with contradictory data because we judge it to be false since it doesn't match what we've previously judged to be true. 

The alternative is that our judgement is off and if that is true, then it calls into question all the other judgements we've made. Our ego feels threatened.

Conclusion (NOT Fact)

All this means that we have to be very careful to not hold tightly to either our experimental conclusions or the base conclusions we call "facts" that those are based on. 

When we are presented with contradictory data, we must be self-aware enough to recognize the potential of our own bias even if we can't quite nail down what effect that bias has. 

We must also accept that the uncomfortable feeling inside us is not necessarily our "gut" telling us something is wrong but could instead be our ego feeling attacked. 

It is only once we develop the self-awareness to recognize how bias drives our decisions and the self-confidence to recognize how our ego drives our decisions that we can effectively harness our analytics to have data drive our decisions

Views: 1953

Tags: data analytics, data effectiveness, data science

Comment

You need to be a member of Data Science Central to add comments!

Join Data Science Central

Comment by Mathieu Landry on November 4, 2016 at 4:29am

Excellent article. Agree 100% with my own biases :-)

Videos

  • Add Videos
  • View All

© 2019   Data Science Central ®   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service