Reinforcement learning is a process of making decisions based on avoiding previous mistakes. As humans, we interact with the world, learning the actions we need to take to achieve our goals. When we learn to ride a bicycle, we learn balance and steering to avoid falling over. In machine learning, reinforcement learning is an algorithm that learns to make the correct decision through trial and error. In the world of project management, we call this experience. Similar to gaining experience, the AI-based algorithm needs historical data. Reinforcement learning algorithms can start with no data and gradually become an expert by learning from mistakes in a game such as chess. However, this may not be the best strategy for managing a project.
Computers can retain a lot of data and have excellent recall. Think of an issue that is captured for a project in progress. What is the problem, and how do we plan to solve it? Project managers gather data and think about possible solutions. We use reinforcement learning in this situation because we avoid a solution that we know failed in the past. Now, think about having a database that contains all the decisions for a similar issue in numerous previous projects. The project manager avoids decisions that do not work and tries a new solution. If the new solution is successful, the reward is feeling good about making the correct decision.
I suggest to my project management students that they start their own project issues database as soon as they are employed in a project role. They can capture the project problem details, the project conditions or environment, the decision made, and if it was successful or not.
|
Project |
Issue |
Project characteristics |
Project environment |
External conditions |
Decision |
Decision success (Y/N) |
|
|
|
|
|
|
|
|
Capturing project decisions is a simple way to create data that an AI algorithm can use to improve project performance. Algorithms use this process by being able to access previous project information to help project managers make better decisions. Imagine if a project manager never made the same mistake twice!
Reinforcement learning is not at the top of the list for AI in project management because supervised and unsupervised learning are easier to work with and provide statistical results. However, this type of algorithm can be a powerful tool for helping project managers make good decisions.




Community Champion