AI Doesn’t Get Leisure . . . or Does It?

Off the Page

By Dr. Jason Page

“As a recreation therapist with a growing familiarity with AI, I initially dismissed its ability to understand leisure—that quintessentially human space between obligations. Yet after witnessing an algorithm generate hundreds of ‘weird recreational pastimes,’ I realized that my own biases were limiting my perspective. While AI may not experience leisure as we do, its ability to reflect our contradictory relationship with downtime suggests it ‘gets’ leisure in ways I hadn’t considered. Perhaps the distinction between artificial comprehension and human experience isn’t as binary as we thought—a realization that has profound implications as these systems continue to evolve alongside us.”—Jason Page

HOMER New York—(Hubris)—August 2025—Let me be clear: I am a fan of AI. Over the past few years, I have been slowly learning various ways to utilize AI as a professional, and, while by no means an expert, I have developed a level of comfort working with these systems. In my professional life, I often use generative programs to help me develop case studies for my students—illuminating the rudimentary elements of our profession as recreation therapists. Beyond this basic use, akin to using a flint to start a fire while leaving the flamethrower to the side, I also enjoy brainstorming new project ideas and working with AI to refine or even completely reimagine project concepts. It was during one such digital dialogue that I thought I had caught AI describing something to me that it could not possibly understand at its core; then I reconsidered.

Over the past five years, AI seems to have morphed into an omnipresent part of our lives. Depending on whom you ask, AI is either the greatest thing since sliced bread or everything that dystopian movie franchises have warned us about. I tend to believe the reality is somewhere in between, but what seems irrefutable is that AI is here to stay and is getting more powerful. 

Some experts in the field believe we are just a few decades from human-level artificial intelligence, this rapid development fueled by increased investment and expanded adoption across a larger segment of the population. This burgeoning inevitably leads to a plethora of ethical questions which are widely debated by policy wonks, industry insiders, academics, and, increasingly, joe public. However, what many users fail to realize is that generative AI transcends mere toolhood—indeed I often hear people referring to it as such, and I too have been guilty of using this simplistic terminology. 

Going back to my early reference, both flint and flamethrower fail to capture the essence of AI since both serve a single purpose. AI is a new form of intelligence, capable of evaluating options and selecting the most appropriate response from its available knowledge. While still capable of making glaring mistakes, AI’s capabilities are far in advance of your word processor’s spell check—a digital tool that has itself been upgraded over the last 30 years. However, this leads me to a further question: what is intelligence?

Intelligence may be simply defined as the capacity to learn, understand, and adapt to situations, often involving abstract thinking, problem-solving, and the ability to apply knowledge effectively. I think I have already made a simple case for AI meeting this requirement. However, humanity is about much more than our ability to process information and while it may certainly be argued that everything we do involves elements of intelligence at one level or another, intelligence is often not the driving force behind our behavior, and this is especially true when it comes to our leisure choices. Indeed, many of the things we choose to do in our free time run counter to what we might otherwise consider effective uses of our time. 

Consider the recreational use of drugs, participating in extreme sports, or even going for a run, each of which involves potential risks to our health. You may think that running is the outlier in my list; however, 70 percent of runners surveyed in one study reported at least one injury in the previous year, many due to overuse. So, when considering leisure, does AI—a system designed to learn and understand even abstract concepts—really understand leisure? After all, it doesn’t seem as though AI is currently capable of “downtime.”

As humans, we are very much capable of “downtime,” when we choose to engage in a vast range of incredibly diverse activities. These activities reflect socio-cultural factors including demographics, geographic location, personal beliefs, and life stage, to name just a few. As a leisure researcher and therapist, I spend much of my time exploring some of the more unusual leisure activities that people engage in and this is what led me here. 

During a brainstorming session during which I asked AI to generate an alphabetical list of “weird recreational pastimes”—I do love my job—I was struck by the sheer range of activities it suggested. As I reviewed the list of 702 leisure activities—created after some back and forth—there were many activities that I found to be outside my understanding of leisure (e.g. Spreadsheet championships for creating the most efficient data systems). My first instinct was to put this down to AI’s lack of a true understanding of leisure, an inability to grasp what leisure really is: that time away from our obligations, and that time’s ability to reflect humanity’s diversity and individual nuance. Then I looked again; this time, rather than dismissing the range of activities the system presented to me as the simple production of an algorithm, I dismissed my own bias. In so doing, I was able to see that while an AI system may not get leisure, it “gets” leisure.

Looking back on the core of my original prompt of listing “weird recreational pastimes,” AI had performed several of its core functions: 1. Identified and alphabetized recreational pastimes (understanding and problem solving); 2. From all the recreation activities the system was aware of, it then identified 702 that it thought would be considered “weird” (abstract thinking and knowledge application). 

It is this second element that I find most compelling. While many of the items on the list may have been available from a range of sources across the internet, many were incredibly obscure (e.g. Queen bee raising or Binary code weaving). This suggests not only the breadth of information available to the system, but also that AI was able to recognize the sheer complexity of human leisure interest, while simultaneously acknowledging the complex array of attitudes and biases we have toward other people’s leisure. While this may seem impressive it should still be noted that leisure continues to maintain an ethnocentric bias toward Western culture and a demographic bias toward white middle-class men. In producing this list, AI simply replicated these pre-existing biases.

Does AI understand leisure? Perhaps not intrinsically, but it certainly mirrors our contemporary leisure landscape. Moreover, given its developmental velocity, perhaps future iterations will engage in their own recreational pursuits. The lingering question then becomes: will AI consider its own leisure weird?

Editor’s Note: Readers following the gestation of AI may be interested in listening to an NPR segment I found illuminating, if sobering: “Who Are the Zizians?” (“On Point/All Things Considered,” WBUR. 30 June 2025)