From Data To Profit

CHAPTER 11The Human-Machine Maturity Model

A primary challenge with AI products is how they interact with people. You might be thinking, “Wait, did he just say that backward?” No. AI products interact with people just as much as people interact with them.

AI products are a novel paradigm for many reasons. The one most relevant to this chapter is their adaptability. As new training data is introduced or feedback is given, they update themselves. Models are designed to retrain and continuously improve.

Retraining a model is the process of introducing new information to the model. In many applicant tracking systems, if a recruiter rejects a candidate who was considered qualified, the weights used to decide what qualified means are changed based on the feedback. These feedback cycles are standard with machine learning–supported features and AI products.

With the original version of ChatGPT, you could correct the model. If it said something wrong, you got an apology, and it updated future responses based on your feedback. This paradigm is entirely different. People are used to adapting to technology. Many are not ready for technology to adapt to them.

In this chapter, I'll cover the most misunderstood aspect of data, analytics, and advanced models. They work with people in new ways, and the implications won't become obvious until models begin to behave unpredictably. Models work exceptionally well while the assumptions they're built on hold. When they fail, there's no warning from ...