Meme Propagation

Deb Roy in his TED presentation on “The Birth of a Word” gives us a glimpse at a technology with potentially high impact.  His primary talk discusses how, over 5 years, with a 18/7 audio/video recording from every room in his house, his MIT team is able to trace the word acquisition of his son. I will let you contemplate the pros and cons of having 100% of your household activities recorded for posterity.

However, his team applied the software they used to capture every use of specific words by his son, then connecting these with every word from members of the household in the proximity of his son, to analyse other interactions.  One source was the feed from every major television network. The second was to track emerging phrases from these sources via the blogosphere/twitterverse.  Their result is the ability to obtain near-real time measurement of the impact that a given source is currently having on the population at large.

A popular TV show may trigger social media flow with a positive feedback loop bringing more viewers into the show.  The proliferating comments may provide analysis of what works best in the show, what is not working, where viewers want the story to go.  One can envision a program driven impromptu by viewer responses measured in the Twitterverse.

However, a second example was President Obama’s State of the Union address. This showed much broader distribution than any single TV show, with massive response and interaction in the Twitterverse.  One can envision real time AI analysis (Deb Roy has been working with Bluefin Labs which does this commercially) that is used to critique a political speech or event.  In the extreme, a presenter may get coaching feedback from real time evaluation, altering the presentation spinning up on the teleprompter in real time.

Consider a political debate where candidates are receiving real time talking points based on analysis of the Blogosphere, and altering their apparent positions based on this.  The good news is that it would require some fairly smart candidates to pull that off, or perhaps ones with nothing in their heads except the words that are being feed to them anyway.  But now the kicker. Activating a zombie army of previously captured devices, all of which have twitter accounts, and you can re-direct the discussion. The discussion migrates from starvation in Iran to education in Latvia. Both candidates arguing the strategic value of Latvia and how their proposals will yield the best educational outcomes.  And the only indication that this discussion has been hijacked is that neither candidate knows where Latvia is. But the good news is that they are supporting education …..somewhere.

Don’t Like This

Hopefully most folks have seen news of the paper in the Proceedings of the National Academy of Sciences that shows how analysis of what you “Like” in Facebook can be used to infer aspects of your personality.  The gist of this is that your “Likes” form a profile about you that can be associated with other aspects of your life — religious affiliation, sexual preference, drug use — with a certain probability.  Presumably other visible aspects of your preferences could also be used in similar ways — who you follow on Twitter, topics in your blog (or on Twitter), etc.  Combining various of these methods is likely to increase the probability that a given “assertion” is accurate.

Some of these things may be “don’t care” for you, but others could be problematic.  With the tendency of employers, schools, and others to evaluate your web presence as part of their interview and other processes, this becomes another subtle channel for discrimination.  Of course the results can yield “false positives”, but it is unlikely that you will know about the uses/abuses of such evaluations and as such won’t have any way to counter the conclusions.

The automated evaluation of many aspects of our digital footprints is something we need to constantly revisit.  “Traffic analysis” is a concept that has been applied to determine communications patterns but also to find the recipe for trade secret foods (three tankers of corn syrup, one of vanilla, etc.)  Watching where you go on the web is something your employer, ISP, search engine, and others can do (it is a key aspect of Facebook “Like” and other such options — and you don’t need to actually select “Like” for Facebook to record your trip to that site.)  Some aspects of this are only visible to the data collector (Facebook) and their friends, but publicly accessible indicators such as “Likes” on your Facebook page are open to analysis by any interested parties.  No doubt we will see emerging services that will either do this on request, or will sort out groups of interest for targeted advertising, or other uses.

As with many possible privacy issues, it is the aggregation of data points that starts to reveal details we might have assumed were at least obfuscated if not private.