[WT] The evaluation of uncertainty in the PP

We have almost clearly defined the goal of the research part of my thesis.

In this post I will try to explain which is our great challenge.

The Planning Poker domain is inherently uncertain as is, in other words there isn’t possible to remove uncertainty in the PP.

This uncertainty derives from:

• the PP itself: depending from the team’s level of experience, from the User Stories peculiarity (a User Story can be ambiguous, incomplete, complex, …)
• the social component: depending from the team members character, from the relations between members, and so on.

This uncertainty could provides a good indicator of the:

• precision of the estimation: more uncertainty implies less precision in the common estimation;
• time needed to play PP: more uncertainty implies more time necessary to reach a common estimation.

There could be a number of research goal such as reduce the uncertainty in the Planning Poker, discover the factors that influence the uncertainty, evaluate the level of uncertainty.

This goals could be very complex and  some of this could be unfeasible, but we decided to start from a the last one: evaluate the level of uncertainty.

In particular, we want to build a tool that evaluates the level of the uncertainty in the Planning Poker. This tool could be useful to:

1. have a measure of the precision of an estimation,
2. stop the PP: when the estimation process takes a long time, the Product Owner have to stop PP and decide the final estimation. PO could use this tool to decide when stop PP (for example when the level of uncertainty go down under a certain threshold)

How can we evaluate the level of uncertainty in the PP?

Combining the uncertainty derived from different source:

• the level of confidence in the own estimate (I’m an expert; I think this User Story is too complex, …);
• the diversity of various estimates (How are different the values of all other member estimates)

Actually this are ideas, and we will investigate more in deep how we can do this.

So far we have found a mathematical theory that combines evidences that could be used for our purpose. This is the Dempster e Shafer Theory.

The theory provides in output an interval delimited by the

• Belief function(b) [0,1]: level of the belief in a fact based on what I know now
• Plausibility function(p) [0,1]: max level of the belief that can be associated in the same fact if all that are unknown should become known

The interval represent the level of uncertainty as in the figure above.

As I wrote this are ideas, and I will provides more details, sample scenarios and all that can be useful to understand in some future post.

Here is my working thesis presentation(in italian)

P.S. Thanks from Adam Porter for the precious advices.

Advertisement

Tags: Adam Porter, Dempster Shafer, great challenge, research, uncertainty