Software engineers at work

I came across this curious university department, called the Software Engineering Observatory.

"researchers can investigate how software developers work in teams, deal with industrial clients and handle the plethora of problems that arise in group projects with tight time-scales"

I expect that an anthropological study of software engineers in their natural environment won't reveal very much of interest, because software itself is metaphysical and the tools required to create it consist primarily of...a lot of thinking. Here's what I imagine that an observer would see if they were to do a time-and-motion study of me.

• Stares at laptop screen
• Taps the same key a few times
• Two hours later, still staring at laptop screen
• Drinks coffee
• Silent, staring out of the window
• Seems to be doing nothing
• Looks like he's praying
• Reading email
  
• Doodles some wavy lines
  
• Staring at laptop screen again
• Hasn't said a single word all morning
• Deletes some code
• Strange patterns appearing on laptop screen - they seem incomprehensible
• Subversion commit (deletions)

Of course sometimes I do actually write code, but probably at least 90% of the time I'm not really doing anything which an observer could directly see or record, other than thinking about the problem and how to solve it. Mechanics and fitters in the past often described me as either "sleeping" or "praying". The remaining 10% of the time usually is conspicuously industrious though.

Even in the most constrained, highly specified environments - such as an industrial production line - the process of developing software is always a creative one comparable to writing a novel, albeit in a very formalised language. Fundamentally it's about engineers creating and implementing solutions to problems.

On a related note I did once read a paper describing the typical eye movements (saccade targets) of users as they read through Google search results. It would be interesting to apply the same measurement techniques to someone writing software, although I think it would still remain difficult to infer from this exactly what is going on at the cognitive level.