http://www.demotivers.com/5412/Who-Doesnt-Belong-Here
- Alexis C. Madrigal, “How Netflix Reverse Engineered Hollywood,” The Atlantic, January 2, 2014
I was struck when reading Madrigal’s article by the phenomenon at the end which he dubbed the “Perry Mason effect.” It instantly made me think of these humor posters about which of these things don’t belong? It was incredible that in a categorization system with literally tens of thousands of genres, that such a strange little hiccup could occur in what one would consider a relatively important category: most popular actors. Plus, this weird occurrence was not linked to recommendations made to Netflix customers, nor did it indicate that tons of people were watching Perry Mason episodes or movies featuring Raymond Burr. In fact, it was just something that happened during the process of using human preferences, fed into a computer, to create these altgenres. There is really no explanation for the Perry Mason effect. Yet when extrapolating this to wider fields in Digital Humanities, I think this occurrence of computational serendipity may be one of the reasons that humanists are so drawn to analyzing their data with machines. The strange feed line of research to computational model or analysis, back to human presentation elucidates incredibly interesting “Perry Mason effects” which the researcher alone would not have seen. However unlike Madrigal, I believe in some cases of research the explanatory reasoning behind the “something in the code and data” can be traced and found incredibly useful by the researcher.
For instance, archaeologists have been feeding information, spatial and quantitative data about artifacts, into databases and mapping programs to show distribution patterns over a whole site or region. Often, nothing strange happens in the translation of the data back to human presentation (the final map for instance), and it shows generally what it was expected to. But in some instances, new spatial relationships, groupings, etc. come to light during this final stage which were not readily apparent, either in the field or straight out of the field notes. Because these computer systems/programs are mechanical, they help the human researcher to investigate the data without our inherent biases and expectations (though those might still be present in the data itself), and let us see things that we would not have otherwise. Usually in these cases, once the “Perry Mason” effect has been identified, it is possible for the archaeologist to retrace how/why/where this might have happened, and to outline something about the site or culture that may otherwise have gone unnoticed.