Need a way to test

I have been listening to Eli Pariser’s “The Filter Bubble”. I became interested in the book after listening to Pariser’s TED talk. I have become fascinated with the claims made in the talk and book.

Here is my take on a core claim of both. The claim might be described as a complaint about either how we use Google search or the system Google uses in proposing links we might explore. The original system, as I understand it, was based on the links among web sites. Linking to another site was thought to represent a vote of confidence in the content appearing on that site and those sites that are the target of many incoming links are regarding as more important in voting for sites they link to. As Google has sought to improve the links we ask for, Google has added data from our click histories. When possible, Google keeps track of what we like and adds our preferences to the criteria used in selecting and ordering the returns to our queries.

It is our personal preferences that seems to concern Pariser. The concern is that the system returns what we want to learn rather than what we need to know. In other words, our existing biases are maintained because we are likely to encounter new information consistent with these biases.

The book explains the basis for this concern which I would describe as a combination of social psychology and cognitive psychology (at least as I interpret the arguments). The cognitive psych component argues we use what we know to interpret new information and we prefer what fits easily (you may remember Piaget’s concepts of assimilation and accommodation from a developmental psychology course). There are important educational implications that describe the challenge of changing flawed ideas. The model is often argued as relevant to science education and is applied to situations in which we have developed faulty ideas based on our observation and misinterpretation of the world around us. The model proposes that we tend to protect our flawed existing models of the world and to create change, the educational system must find an effective way to activate and show the faults of existing ideas. Both activation and perceived inadequacy are important or the flawed ideas will be retained.

So, translated into the argument Pairser makes – if we have flawed ideas, we are likely to also have prioritized information sources that support these views and the history of the clicks then continues to feed these flawed ideas rather than present information that will lead to change.

Anyway, I have trying to evaluate the flawed search argument without much luck. Pariser describes a recent situation in which two acquaintances of his received different links when searching for Egypt (following the uprising). The claim was that these two individuals had different interests which may have resulted in different sites being prioritized. So, I have been trying to create my own method for generating different link lists.

My strategy to generate an equivalent demonstration has taken the following approach:
1) I use Google for many activities and log in (so it knows who I am and could connect this with my email content, my docs content, my musical preferences, and my searches). I assume that my content is heavily tech oriented and this could be considered a bias.
2) I have tried searches focused on the words “apple” and “chrome”. These are terms with multiple associates one of which involved technology.
3) I have tried signing in before search and comparing these results with the results generated when I do not sign in.
4) I have tried comparing the results from my Mac (used a lot) with results from a Windows machine (used very little).
5) I have tried comparing the results that can be identified with my local ISP with results through my mifi (not associated with my local ISP).

So, the strategies above are intended to include or not include data that can be linked to me.

None of the searches for either term seems to generate different results. All results are tech dominated. Nothing about apple pie or embellishments for hot rods.

This approach seems sound, but I need different search terms. I think I need terms for which my meaning would be in the minority when contrasted with what most people would be looking for. I am looking for suggestions.

Powered by ScribeFire.