The diminishing returns of too much information

I really liked this image of the inverted U from Peter Morville’s Ambient Findability so I decided to make my own version:

I’m a strong proponent of data-driven practice, particularly for classroom- and student-level progress monitoring of essential learning outcomes. But, as Morville notes, there are diminishing (and even negligible) returns if one goes too far. We have to know when to say when.

4 Responses to “The diminishing returns of too much information”

  1. I think this is a great point, but I am going to suggest a minor change. I think the label on the X axis should read “Amount of Information Under Consideration” instead of just “Amount of Information.” I like your way of putting it better than Morville’s “Volume of Information,” but I still think the essential part is “under consideration,” not the “volume” or “amount”.

    I don’t see any problem whatsoever with increased volume of information. The more we know the better – if the information just sits there and is never used, so be it, it is not harming anything. For instance, I don’t think that a larger dataset with 50 columns is any worse than a smaller dataset with 20 columns of information. The trick is to know which columns are relevant and which are not. But, we shouldn’t use rationale as a justification to limit information production, we should use this as a justification to get better at information management.

  2. Maybe, Justin, but there’s also the ‘signal-to-noise’ ratio to consider. At some point, filtering through the excess (e.g., my Internet browser bookmarks!) to get to what you want becomes too difficult and/or time-consuming. Why dig through 50 columns when you only want 6?

    That said, you can add whatever qualifiers you want to the graph – that’s okay with me!

  3. Very helpful visualization, thanks. Although the issue may well be semantic, I wonder how data quality – rather than simply quantity – affects this graph. What I take from this is despite (maybe because of?) a phenomenological argument like Morville’s seems to be, what I take away is an increasing need for information literacy education rather than simply letting our students ‘become what they find.’

  4. Scott, I like your visual of the inverted U. Information is excessive if it’s not somehow pertinent to the goal or if it’s relevant but can be consolidated into a more readable and therefor more useful form. Almost anything can be quantified, (for example, the number of cells lost from my fingertips during this exercise!)but unless it ultimately relates to the chosen topic, that data is noise. Data teams need to consistently review the relevance of their data and the targets of the measurement to avoid “data overload” or worse, drawing conclusions from irrelevant noise.

Leave a Reply to Mike Hetherington