Cognitive Offload

A recent study from CMU and MSFT has been making the rounds all over LinkedIn. The study finds there is a risk of over-reliance on Generative AI, which may lead to diminished independent problem-solving skills among knowledge workers. The exact type of problem-solving skills that Generative AI was supposed to give us more time to do because Generative AI was going to take care of the monotonous, repetitive drudgery of our day-to-day jobs.?

The funny part of Enterprise Software is that people use tools for all sorts of unintended purposes…but we all should have seen this coming.?

Foreseeable Outcomes

A workflow that led to people using Generative AI for cognitive problem-solving looks something like this:?

1. Generative AI is applied to rote tasks and increases productivity
2. Increased productivity results in reduced headcount
3. Reduced headcount drives the need for existing resources (people) to take on new responsibilities?
4. New Responsibilities are unfamiliar, and the people who got laid off didn’t train their replacements
5. Generative AI is applied to new tasks to increase productivity?

What are the challenges with this workflow other than its human impact?

Quality slips because the ‘fresh perspective’ that new people bring to old problems is gone. It is now replaced with decisions made by data that isn’t refreshed but recycled. Solutions are commoditized.?

Datatuition?

Humans make decisions based on a combination of data and intuition. Generative AI can easily manage the data collection part of the problem. With enough data it can suggest a solution that it predicts based on an average it has retrieved. It makes a suggestion even if it doesn’t have enough data because it has a job to do, which is to answer your questions.?

Generative AI doesn’t have intuition because it doesn’t have enough context. At least not yet. Lots of really smart (like, stupid smaht) people are working on creating world models, which are complete views of the environment we occupy via sensors and super-blazing-fast chipset technology. So, intuition is coming, but it’s not here yet. Yet..Generative AI is being used to make decisions that are better made by critical thinking. Or so we thought.?

Silver Bullet Solutions

Apply the cold, logical observation of human behavior via an old maxim: If you have a hammer, everything looks like a nail. You can use a hammer to put a screw into a wall, but it doesn’t work the way it’s intended to. Yet, the screw technically solved the problem if someone miswrote the problem statement. If you leverage technology the wrong way but still get the job done and the difficulties are….someone else’s to deal with, what’s the actual problem??

Learning is the problem. We don’t learn how to apply logic in another context without previously thinking and coming up with solutions. Using Generative AI as a shortcut results in knowing how to use a tool better, which is not a unique skill.?

Emerging Disruption?

AI is a technology being perfected as it’s being used. In the process of using it, people were simply testing it out and applying it to all sorts of challenges. And it worked because sometimes there are a finite number of viable, feasible, and desirable solutions. It also worked because it’s predictive, and some problems can be addressed by going through the combinations and permutations of approaches. The reality is that a machine can do detail-oriented, data-intense work with limited human oversight.

Certain types of problems may not have required as much critical thinking and reasoning as we previously thought. But solving a problem often differs from creating a scalable, sustainable solution—that type of problem-solving requires reframing a situation. Re-framing is more of an art than a science.?

Critical reasoning and creative problem-solving are valuable skills. Without them, we are easily replaced.?

Failing to prepare and preparing to fail?

Most companies that bought licenses to use Generative AI did not spend the time enabling or training staff on how to use the technology. In most cases that software is sitting there, collecting dust, waiting for a job, in fact, 82% of enterprise companies claim to be training AI, but only 40% have deployed anything to staff. There has been disproportionate adoption of the technology, and without guidance or leadership modeling desired behavior, most people aren’t utilizing it the way it was meant to be. But in any system, we have outliers.?

It’s not shocking that early adopters remain early adopters when applying the technology in different contexts. The first person you knew using ChatGPT to plan a vacation is likely using it to do something you hadn’t thought of today. Ask them about it.?

Individuals who use Generative AI more than others are still humans. They may not be sharing everything they learn because it gives them a competitive advantage, or maybe they are concerned about how much of their job they can automate. Without an incentive to share how the tool can be used, it won’t be used as often.?

If an organization doesn’t provide guidance on technology use, you don’t need an LLM to predict unintended use.?

Solving the wicked problem?

Providing guidance and ongoing support to staff on the how and when of AI is a good start to long-term adoption and effective deployment. Assessing what types of problems require critical reasoning and creative problem-solving is worthwhile, as there are human methods of achieving efficiency that the application of AI will not improve.

At Gyroscope, we partner with you to support your adoption and implementation of intelligent automation. We provide education and enablement services to increase the utility of tools and facilitate workshops to identify measurable opportunities to use AI. If you would like to get started today, please reach out to us here.