Filtering information and seeing patterns that may not be there

The difference between correlation and causation is well known among most data analyst folks, as is the risk of drawing a false conclusion about a causal connection between two variables that are correlated but not causally related. I was talking with my husband last week about the increased volume of data and information we are exposed to with the expansion of information technology in our daily lives, and the resulting increase in likelihood that we draw connections between two events that may not be connected. (Nate Silver does a fantastic job of discussing this outcome of larger amounts of data and need to adjust statistical modeling techniques in his book The Signal and the Noise.)  The conversation got me thinking about how we filter information. We choose what we pay attention to, and while we might be able to expand our ability to take in information somewhat, I think there is probably an upper limit on the information we can take in. Which means that as more information comes at us, we filter more out.

We are creatures that make meaning out of information. It seems to be a fundamental part of being human. Pattern recognition is wired into our brain structures and processes. In Buddha’s Brain, Rick Hanson explains the neurobiology at work: our ability to quickly make meaning of data, to recognize threats and respond, significantly increased our chances of survival. Similarly, our ability to project into the future helped us survive with our ability to plan for the worst case possible event and try to avoid it. One way we make meaning from data is pattern recognition. We do this by matching incoming data to patterns we have already identified. One essential meaning-making skill we developed as humans was with facial recognition. And interestingly, there are two aspects of facial recognition that operate differently but have the same ultimate purpose of increasing survival.   First, we are very good at recognizing faces of other people and categorizing them into groups – primarily “us” vs. “them.” This served us because recognizing those who were not part of our group increased the chances of group survival, when we were in competition with other groups (which was for a very long time in our history). The other aspect of facial recognition relates to our skill in identifying the emotional state of another person – this increased our ability to stay attuned to others within our group, and thus our ability to get along with them, which increased our survival chances as individuals, since it would have been very difficult to survive without group membership.

However, in optimizing our filtering of data and meaning making for survival, we have a tendency to disregard a lot of information that could be useful for learning or longer term success, especially if it challenges our current world view, and at the same time to assign meaning to information that is not actually meaningful. We tend to pay more attention to data that reinforces our mental models, and filter out those that refute them. This shows up in a lot of interesting ways: in scientific practice, it can take a change in generation before a new theory is recognized as more valid (see Thomas Kuhn’s The Structure of Scientific Revolutions); in forecasting, analysts can “overfit” a model, assigning significance to data that is actually just noise (again, see Nate Silver’s book); in interpersonal relationships we often run up the “ladder of inference” in seconds – starting with selecting data from what we observe about other people or a situation, adding meaning, making assumptions, drawing conclusions and adopting beliefs, before taking actions – and our beliefs affect what data we select (Chris Argyris first developed the ladder of influence concept).

I don’t think it is possible for us to avoid filtering or assigning significance when it does not exist. It is hard to be truly open to new information that does not fit what we already think. But in open dialogue with others, where we balance advocacy and inquiry, we can collectively come to a deeper understanding of information. Group learning can balance individual entrenchment. To practice this, we need to practice being open to each other. For more on balancing inquiry and advocacy, see chapter with similar title by Rick Ross and Charlotte Roberts in the Fifth Discipline Fieldbook.

Share this postPrint this pageEmail this to someoneTweet about this on TwitterShare on LinkedIn
This entry was posted in Uncategorized and tagged , . Bookmark the permalink.

One Response to Filtering information and seeing patterns that may not be there

  1. Carlos A. Ochoa says:

    Dear Eleanor,

    Thanks a lot for sharing your great post about how to manage the noise factor when working with big amountsof data, in my very humble experience as Global Advance Quality Engineer at General Motors I worked with hundred of thousands of warranty and customer survey verbatims, trying to identify trends of failure modes, customer usage patterns, preferences,etc.

    To reduce the noise factor I usually used the DFSS P diagram tool, where I can state the input factors, control factors, define the response or ideal function, error states or non ideal functions and noise factors and after that create a blog diagram to find all the relationships possible between subsystems or components.

    By first hand I know the importance of make the right call in trends or patterns, and the pain of wrong conclusions that you can get if you are not paying attention to the noise factors, in my case were giving wrong directions for future designs.

    After 20 years of auto industry experience I am giving me the chance to find or create my “ideal” job which combine data analysis focused in a phylantropy goal, helping others is part of my ADN, as Rev. Martin Luther King once said:

    “Life’s most persintent and urgent question is what are you doing for others.”

    I would like to have the opportunity to exchange some mails with you, I would really appreciate your advice and directions about how to combine these 2 passions.



Leave a Reply

Your email address will not be published. Required fields are marked *