Papers on agile can be found at George Dinwiddle's WIKI on this topic. Lots of good reading here, but some troubling approaches. First let me give my point of view. I'm from a very biased training and experience background - physics and more importantly particle physics. In the late 70's (my era) there was lots of speculation about the existence of certain particles and the forces that cam with them. In the Standard Model and its derivatives of today, forces between particles and in fact forces between anything is carried by other particles. The electromagnetic force is carried by photons. The signal received by the antenna of your car from the local NPR station is "carried" to the antenna from the transmitter by photons. The electromagnetic field theory of electrical engineering works wonderfully to design and build the FM transmitter, it antenna, the car antenna and the car receiver. But at the particle physics level, the "field" disappear and particles take over. This how both waves and particle can exist.
Anyway, the process of experimenting - my domain, since I was not smart enough to be a theorist - is based on a simple protocol of talking about ideas and the measurements associated with those ideas. One process of this measurement is the statistical assessment of correlation. Correlation between tow variables. Say a setting in the experiment and the measured outcome of another variable.
The correlation coefficient states how the two variables are related. Low correlation coefficients - say 0.1 - mean there is only a 1 in 10 chance the tow variances have any connection to each other. A high correlation coefficient - say 0.9 - means there is a 9 in 10 chance the two variables are correlated.
This comes into play in every discussion about cause and effect. I walk into the thesis advisor's office and say "hey I just saw some numbers come from my computer sampling system (this is how I learned to program computer, we were called programmers back then)." "I think we have some interesting data to show you." The first or second question asked by the Research Associate, who was probably a few years ahead of us, would be: "why should you see these numbers?" and "how are these number correlated with the setup of your experiment?" Good questions.
These two questions are related to two important experimental processes
- Hypothesis testing
- Correlation of the variable
Are the Agile Papers Science?
The question comes to mind when I read most these papers that they are first anecdotal and second when there is a controlled experiment both the controls and the correlations factors are weak at beat. Weak from a physical sciences point of view. Remember I'm biased here. They appear more like social science papers.
And that's the point I guess. When there is a study of Agile Effectiveness or Measure of Agile processes - they are social science class measurements. One side hypothesis tests, with weak correlations. That's fine. It's anecdotally based social science. You can get a PhD in social science, just like you can in Finance. But these disciplines all have different standards for what "research methods" are used to validate the results of experiments.
There an interesting passage in the new Walter Isaacson Einstein biography. In the 1912 "notebook" Einstein proposed a two-fisted approach to General Relativity - a physical strategy and a mathematical strategy. This approach - the two fisted-kind - can be applied to project management and agile project management processes
- Physical - the observational evidence that things are working or not working in a controlled environment. By observation I mean that processes are observed, measurements made between "before and after," or "control group and agile group."
- Mathematical - the measurement basis of the processes of project management. Either statistical - Monte Carlo Simulation, or deterministic - Earned Value.
This "may" an approach to join the observational with the process based project management methods.
Next I'll finish reading all the papers and try and construct a framework for this approach.
