AI IQ

A new field of study known as explainable AI seeks to make AI results easier to understand. A problem with algorithms using regression analysis on large datasets is knowing how the results were obtained.  Neural network algorithms adjust weights internally to achieve the highest correlation to determine the results. The issue is understanding which characteristics of the datasets are most significant.

For example, AI software might indicate that a risk is missing from the risk plan, more training is required for two resources, or a specific change request is highly probable. A stakeholder asks why, and the explanation is not readily available. Explainable AI is a field of study that seeks methods to determine how AI works and how to make the results more transparent.

Here are some methods I use to analyze AI results.

1. Manipulate the features. This is similar to performing a simulation but running the algorithm several times with slightly different data to determine what data is most relevant to the results.
2. The number of datasets used in project management to make predictions is usually smaller than in other disciplines. For smaller datasets, I use a genetic algorithm to determine the data features that have a higher impact on the results. Typically, the software shows several features and indicates the factors that account for over 90 percent of the output.
3. The last possibility is to change the project datasets by adding, reducing, or using different datasets. This is similar to my first point but requires a focus on complete datasets instead of data fields.

Why does it matter?

There are several reasons to be concerned about AI delivering results that project managers don’t understand. One of the reasons is ethical issues such as biased data and determining how bias affects the results. As AI becomes ingrained in project decision-making, explainable AI is required to detect and explain abnormal results. Academics and programmers continue to progress with different approaches, such as measuring the level of trust in the result.

Conclusion

As project managers learn to collaborate with AI software, explainable AI will play a critical role in building confidence that decisions can be taken based on AI results.

References

Hoffman, R. R., Mueller, S. T., Klein, G., & Litman, J. (2018). Metrics for Explainable AI: Challenges and Prospects.

Feiyu Xu, Jun Zhu, Wei Fan, Yangzhou Du, Hans Uszkoreit, & Dongyan Zhao. (2019). Explainable AI: A Brief Survey on History, Research Areas, Approaches and Challenges.

Have you ever finished a project that did not go well and looked back at all the issues and surprises? It would be valuable to know them before the project begins. Prediction is one of the main capabilities of AI software. The objective of using AI is to predict all those surprises and difficult issues before the project begins and allow project managers to evaluate them. A good project manager is proactive. In other words, we want to take action to prevent serious problems before they occur. It's not always possible, but even knowing about them in advance can help prepare the project manager for what to expect.

We already predict many everyday events.  We know that if we go outside when it is raining, we will get wet. We know if we drive through a red light, there might be serious consequences. Predictions are made without much thought because of our experience and knowledge. AI takes prediction to a higher level and makes predictions based on project environmental data that a project manager cannot discern. Projects can be complex, with many factors that are difficult to evaluate.  

Predictions based on AI software use three main methods to predict what will happen on the project.

Supervised learning is based on historical data. The software compares an image of your project to previous successful projects to determine if the same adverse events are likely to happen. The prediction can cover all aspects of managing a project, including resource issues, risks, cost increases, schedule delays, communication problems, quality issues, and unexpected interactions with other projects.

Unsupervised learning does not need historical data. The AI-based software clusters or groups items by project based on their characteristics. The similarity in groupings is used to make helpful project predictions. Risks can be grouped to determine if there is a common cause or if they can be managed with a similar risk response plan. Tasks can be grouped to investigate the level of complexity. Resources can be grouped to assess the variety of technical skills. The objective is to predict the potential for project issues in these areas quickly.

Reinforcement learning relies on events from previous similar projects. This is similar to having a lot of project experience, but the software remembers it all. When issues arise, the software predicts which actions will be ineffective. It might offer a solution to resolve the issue if one is available. If not, at least you know what will likely not work and can look for alternative solutions.

The concept of predictions is to move knowledge of what happened during the project to the beginning of the project or before a critical decision is made so the project manager can decide how to proceed. Several AI software vendors have this capability, and it is a lack of awareness that prevents organizations from taking advantage of the powerful opportunity of prediction.

Genetic algorithms are a software representation of the theory of evolution and can be useful in solving many significant and different issues in project management. Humans reproduce, and the offspring tend to look similar and have similar—but not identical—traits to their parents. There is no typical pattern to this process. Offspring might receive 80 percent of their genes from one parent and 20 percent from the other, or they may receive 65 percent from one and 35 percent from the other. In a further twist of nature, a random percentage can be included, known as a mutation. As with the development of any species, those most able to adapt to the environment survive. This survival concept is known as a fitness factor and is important in the project setting because it is used to represent the project objectives. The genetic algorithm simulates evolution by creating all possible combinations of a solution until the one closest to the desired result is found. 

How is this used in project management? In projects, the objectives are known, typically the scope, budget, and scheduled end date. Given the desired result, the algorithm searches for all possible combinations to achieve the objective. The value is that a genetic algorithm is not constrained by human bias, knowledge, or experience. The algorithm churns through all possible combinations of potential decisions, including solutions that a human project manager may never consider.

A review of research on using genetic algorithms in project management reveals that project scheduling is the most significant opportunity (Ancveire & Poļaka, 2019). The studies describe results for resolving scheduling conflicts, optimizing resource leveling, developing a scheduling strategy, improving critical path planning, and determining solutions to resource constraints.

This is an example where AI-based algorithms can solve numerous project issues. Collaboration is required between project managers and software developers to find ways to unleash the power of genetic algorithms to significantly improve project performance. These types of algorithms can be more complex to understand but are an example of another wave of machine learning that is available to solve project issues.

 Reference

Ancveire, I., & Poļaka, I. (2019). Application of Genetic Algorithms for Decision-Making in Project Management: A Literature Review. Information Technology & Management Science, 22, 22–31.

I recently returned to Canada from teaching my course, Applying AI to Project and Program Management, at two business colleges in France. This is the fourth time I delivered this course. Students at the Lille and Paris campuses have an option to take the course as part of their Master’s Degree in PPM. None of them had any background in AI, and they were excited to learn how organizations use the different components of AI to improve project performance.  

Although many students are from France, the international representation included Greece, Germany, Poland, China, India, the Middle East, and one fellow Canadian who is an exchange student. While there was a lack of knowledge about machine learning, all students had already used ChatGPT, which is a good indication of their interest in innovative technologies.

The course covered the basics of using algorithms for prediction and classification. I also covered natural language processing and the evolution of large language models (LLMs). Since the students were unaware of AI software vendors that have solutions for project management, we reviewed six interesting ones and how the capability is currently being used by some very forward-thinking companies. All of the AI software vendors offer free demos.

There are now 60 additional students who understand how AI is changing the way we manage projects. I encouraged them to include the course and any new insight about AI on their resumes. Businesses that want to bring that new knowledge to their project environment should consider hiring them. If you have an employee who wants to define a project data strategy and discuss ways for your projects to take advantage of AI technology, they have probably completed my course. 

Cognitive bias is the human tendency to reach decisions not based on logic. In project management, optimism bias is a significant problem when determining the accuracy of a project budget and schedule. Research by Nobel prize winner Daniel Kahneman challenges the belief that humans make rational economic decisions. He proposed the term planning fallacy to describe how we underestimate task completion time and cost regardless of available historical estimates for those tasks. Optimism bias is a significant and pervasive issue in project management.

Professor Bent Flyvbjerg expands on Kahneman's work, describing ten behavioral biases common to project managers. Common project estimating techniques include analogous, parametric, template, simulation, and function points. Despite the systematic methods and probability calculations, the ongoing evidence of budget overspending and schedule delays suggest that traditional methods are ineffective at eliminating optimism bias.

Academic studies reveal that machine learning algorithms are more accurate at forecasting project budgets and project duration. AI has the potential to eliminate or dramatically reduce the bias that human project managers include in the estimating process. Machine learning algorithms can develop a bias based on collected data, but the bias can be detected and removed. Human bias tends to be more ingrained. For project performance to improve, organizations must move beyond existing methods and adopt AI technology.   

For further insight, here are a few references.

Flyvbjerg, B. (2022). Top ten behavioral biases in project management: An overview. Project Management Journal 52(6). doi.org/10.1177/87569728211049046

Ghimire, P., Pokharel, S., Kim, K., & Barutha, P. (2023). Machine learning-based prediction models for budget forecast in capital construction, CEES 2023 2nd International Conference on Construction, Energy, Environment & Sustainability.

Kahneman, D., Rosenfield, A. M., Gandhi, L., & Blaser, T. (2016). Noise: How to overcome the high, hidden cost of inconsistent decision making. Harvard Business Review, 94(10), 38–46.

Min, A. (2023). Artificial intelligence and bias: Challenges, implications, and remedies. Journal of Social Research, 2(11), 3808–3817. https://doi.org/10.55324/josr.v2i11.1477