CMMI-P-SPEM Experiment

We would like to invite you to participate in the experiment where we will explore novel software modeling techniques.

Recent studies (the Babel Experiment, http://www.vlpavlov.com/ooaood) indicate that teams of software designers are capable to create quality models even when use of natural language (spoken or written) is forbidden and designers are allowed to communicate using only Unified Modeling Language. There are reasons to believe that prohibition to use natural languages can make design teams be more productive in some cases (see details below). To validate this hypnotizes and evaluate how “speechless” approach can be used to improve effectiveness of existing design techniques, we are going to conduct a new experiment, called CMMI-P-SPEM (below we explain why we name the experiment in this way).

This experiment will be organized during Software Engineering Conference (Russia) on October 26-28. All participants of the CMMI-P-SPEM experiment will get free full access to all the data, collected during the experiment (including created models), and a free license to use these data in their work. If novel modeling technique will be defined, all experiment participants as well as their current employers will be granted free licenses to use this technique in their work.

Below we provide detailed information about:

Previous experiments (the Babel Experiments);
The upcoming CMMI-P-SPEM experiment;
Schedule and registration.

The Babel Experiments (the Previous Experiments)

The Babel Experiment was designed in 2001 by Vladimir L Pavlov and has already been repeated more than ten times. The essence of the Babel Experiment is that a team of participants is supposed to create high-level design for a software system. They have several hours to complete the task. During this timeframe verbal and written communication using English/French/German/Russian/Ukrainian or any other human language is forbidden, and the UML is the only allowed language. So, experiment participants are to discover whether the UML is “a real language” that can serve as a communication tool to help a project team to achieve the common goal. The Babel Experiment has been repeated more than 10 times and it has always been successful – teams were always able to complete their assignments. You can find more information about the Babel Experiment at http://www.vlpavlov.com/ooaood

The Babel Experiment was initially designed as training, intended to make students:

go through the communication problems that are typical for large software development projects;
obtain the successful experience of applying UML to overcome these problems.

In several instances, the experiment was carried out simultaneously for two teams with the same design task. One team, Speechless, was allowed to use only the UML and pantomime while another, Chatter, was allowed to talk in addition to using UML. At the end of the experiment Speechless and Chatter models were compared. Results of comparison have shown that Speechless produce more mature and detailed models. This lead the author to develop new ground for a more effective modeling techniques based on the following premises:

Speechless members stay focused on their assignment - they can not spend time on jokes or discussions about recent news;
Speechless members tend to explicitly specify (uncover) all details and assumptions; and
Speechless members care about making their models "readable".

The CMMI-P-SPEM Experiment (the Upcoming Experiment)

As it was mentioned above, the Babel Experiment was designed to provide students with new experience. The new experiment, CMMI-P-SPEM, is designed to:

validate Vladimir L Pavlov’s hypothesis that “speechless” approach to organize work of design teams in some cases could be more effective than “regular” approach; and, if so
define a new modeling technique that would be based on “speechless” approach.

It is called CMMI-P-SPEM because of the nature of assignments that will be used during this event – experiment participants will be required to model some of the CMMI process areas using UML and SPEM.

Unlike the Babel Experiment, the CMMI-P-SPEM is indented for experienced professionals, not for students. So, to participate in this experiment one has to meet the following eligibility criteria:

over two years of experience in creating UML and/or SPEM models (reference letter form the current or past employer is required as a proof of experience); and
over two years of experience in applying CMM and/or CMMI in practical/research work (reference letter form the current or past employer is required as a proof of experience).

All participants of the CMMI-P-SPEM experiment will get free full access to all the data, collected during the experiment (including created models), and a free license to use these data in their work. If the experiment is successful, and a new modeling technique will be defined, all experiment participants as well as their current employers will get free unlimited licenses to use the new modeling technique in their work.

One must register to SEC(R) conference (http://2005.secrus.org/) in order to be able to register to the CMMI-P-SPEM experiment.

CMMI-P-SPEM Cost Schedule

Registration date Participation in SEC(R) only Participation in both experiment CMMI-P-SPEM and SEC(R) conference

Before September 31st $150 $250

After October 1st $200 $350

Tentative Experiment Schedule

August 15, 2005:

Organizing committee sends draft rules of the experiment to registered participants;

September 1, 2005:

Deadline for submitting feedback about rules of the experiment;

September 19, 2005:

Organizing committee sends final rules of the experiment to registered participants;

October 26, 2005:

“Introduction to SPEM” lecture for those, who have experience with UML but is not familiar with SPEM (2 hours);
Games (ice-breaking, charades, etc. – 2 hours);
Parallel design sessions (5 hours);

October 27, 2005 (4 hours):

Workshop to analyze results of design sessions and define the new modeling technique;

October 28, 2005 (1 hour):

Presentation about experiment results for SEC(R) participants.