Ethnomethodology is neat

On Wednesday I attended a talk given by Bob Moore (Senior Research Scientist at Yahoo).  He is a former Xerox PARC researcher, and currently does the bulk of Yahoo's exploratory research on internet experiences.  He used an ethnomethodological approach, and conversation analysis to explore how a single user interacts with a computer, specifically using search engines.  Ethnomethodology seeks to find accountable actions and interactions in "doing something."  Rather than seeking to interpret the particular event, it attempts to describe the sequence of orderly actions that take place to complete "doing something".  This approach requires more rigor in data capture (everything must be recorded), and analysis (conversation analysis is often done, which requires transcription and coding). 

Ethnomethodology has strong roots in user research, and many argue the ethnographic practices used today in the field are an unneccessary shift away from ethnomethodology--as computers moved outside of the workplace and into the home, researchers became more interested in "culture" and "society" rather than situated action.  As a result, many researchers felt that a new research methodology was needed (which is the more hermeneutic and interpretive form we see today).  There is a controversial paper that discusses this deviation from ethnomethodology by Crabtree, et. al, Ethnography Considered Harmful.  If you're interested in what that's all about, let me know and I can make another blog post.

Moore provided compelling arguments for the use of ethnomethodology in user research outside of the workplace, and also in single user interactions not relegated to social settings.  Using search engines, he argued, is very simliar to human interactions in which a service is provided.  The quality of service provided depends on a number of factors:  the knowledge of the service provider, the knowledge of the service receiver, and their referential connections.  He gave the example of a simple copy shop interaction: 

A man walks into a copy shop and tells the clerk he wants 10 copies of a sheet of paper, and he wants them "coated."  She replies "what's coated?" He gestures with his hands indicating two sheets of paper coming together and says "ya know, when the two plastic deals cover the sheet."  She then replies, "oh, laminated?"  He affirms her response.

In this example, there is a problem with the point of reference (something is mis-named).  The clerk and the customer both attempt to repair the conversation using names, gestures when only names fail, descriptive wording, and questions.  Moore feels that there are a lot of similiarities between the types of repair in these conversations, and search engine solicitation.  He had his participants perform a series of tasks: he provided images with a caption that participants needed to search.  For example, he would provide this picture with a task:


FIND THE CLOSEST HOTEL TO THIS PLACE.

There were a few ways that participants would search:  by the entity name (as if the copy shop customer were to say to the clerk, "laminate"), by a related name (the customer moving his hands together to indicate lamination), a generic description (the customer saying "two plastic deals").  He found that when a generic description is used, for example, with finding hotels near this place, "monolith that looks like a tree and mountain", that it takes an average of 6.7 repairs to get to the desired response.  Repairs in online searching can either be user-initiated, such as retyping, deleting, etc. or it can be initiated by the search algorithms, such as "did you mean..." or auto-complete (the clerk asking "what's coated?")

From this research, he determined that the goal of an online search is to refine and identify a proper name of something (hence, why the participant didn't even mention hotel in the original search, they were trying to find out what "this place" was called).  This name-centricity, he claims can drive future search algorithms--perhaps yielded results can offer websites that offer more proper names, such as Wikipedia.  He noticed too that participants often referred to the image tab to identify a name of something.  Therefore:  a descriptive search term + a reference to the images tab can mean "I'm having trouble finding the name of something."

I felt that his talk was very useful in understanding the application of ethnomethodology outside the context of workplace studies, and in the context of solitary user interactions with something.  I felt that his protocol--particularly only using images and vague text for his tasks may have biased the opportunities he defined (going to images=trouble), but overall it was interesting.