Observations on Knowledge Transfer of Professional Software Developers during Pair Programming — Franz Zieris

Franz want to understand how knowledge transfer works in pair programming., and then use to do define patterns and best practices. For that, we need data! Franz asked companies to be allowed to record screens and audio during real programing activity. Franz now has 50 sessions from 11 companies. A perspective 🙂

2016-05-20 12.03.31

He selected 13 sessions that contained 450 knowledge transfer moments. This is qualitative research, so those were the last numbers, says Franz.

A few examples.

2016-05-20 12.05.35

What happens here? Bob has an internal knowledge need, he needs a missing puzzle piece. Carl has what Franz calls the target content (the missing piece) This is a pull scenario, and their are also push episodes where Carl would estimate the need and just start explaining.

This was a very small scenario, and push and pull scenarios can be combined, in a bigger episode where cocreation is happening:

2016-05-20 12.09.25

It is not always a happy path, there are problematic paths too, if there is no resynchronization. In this example, the problem is the 40 minute struggle:

2016-05-20 12.11.22

To summarize:

  • Some knowledge transfer episodes can be quick
  • But only if successful synchronization happens
  • Even worse, the pair can end up on parallel tracks.

Other antipatterns include the talking pioneer.

Preprint is here: http://www.inf.fu-berlin.de/inst/ag-se/pubs/ZiePre16-ppknowtrans2.pdf