I TRIED OUTSOURCING MY DECISIONS TO CHATGPT AND ALL I GOT WAS THIS INCREASED EMPATHY FOR THE HUMAN CONDITION

Many aspects of wellness seem to be about making healthy decisions: choosing to eat a salad instead of a pizza, say, or writing in your gratitude journal instead of texting your ex. But decision fatigue can become quite draining, especially when your own overall wellness seems to be on the line. I’ll admit I’ve occasionally wished to relieve myself of the burdens of autonomy and take a break from choosing altogether.

So one night, while eating a pizza and waiting for my ex to text back, I had an idea. Maybe ChatGPT—the buzzy new tech possibly responsible for the sudden increase in my composition students’ verbosity—could make my wellness choices for me. And maybe, unlike my therapist, it would be willing and able to do so 24/7, at a moment’s notice, and without asking about my childhood first. I’d try it for a day, outsourcing as many potentially healthy decisions as possible to the AI. No more ruminating, no more waffling. Easy, right?

Well, sort of. As it turned out, getting ChatGPT to actually tell me what to do was harder than I’d imagined. The AI resists giving direct commands, even if you assure it that you “just want to be bossed around for fun.” Its default mode with simpler questions—what to eat for breakfast, for example—was to spit out a list of general guidelines or healthy options such as yogurt with granola and berries (“This is a quick and easy option”) or a smoothie bowl (“This is a refreshing and colorful option”). But it also tended to hedge that “it all depended” on other variables like the contents of my fridge, my hunger level, and my dietary restrictions. In the time it took to type those things out, I’d lost my appetite for AI-assisted meal planning. Asking the bot what time I should wake up in the morning resulted in similar tedium.

Besides, the decisions that really exhaust me aren’t ultimately about oatmeal vs. eggs. When I asked ChatGPT whether I should end a complicated romantic entanglement, it politely demurred (“As an AI language model, I cannot make that decision for you”). When I asked if I should go visit family next weekend, its response included lines as guilt-inducing as if I’d asked my own mother (“Consider your mother’s feelings, and whether your visit would be helpful and beneficial to her and the rest of your family”). No matter how I rejiggered my prompts, it stubbornly stuck to providing lists of factors that “may be helpful to consider” when making my own (*$&^%$) decision.

As the day wore on, a different kind of fatigue began to set in. At this point, I had forgotten what regular unassisted thinking felt like, so of course I asked ChatGPT to help me describe my state. It suggested:

I began to feel a sense of existential dread. Who was I, if not the master of my own destiny? If I let a robot make all of my decisions, was I still truly alive?

Okay, that was a little dramatic, especially considering ChatGPT’s own refusal to master my destiny for me. Really, I was just tired of trying to get the AI to tell me what to do. But this, I realized, was its own kind of insight. Prompting and re-prompting ChatGPT felt a lot like rumination: I was just going around in circles with the bot instead of myself. It was interesting to see this process externalized: All that typing and scrolling made it more obvious how exhausting it was. Why was I putting myself—not to mention ChatGPT—through any of this?

In a way, the difficulty of outsourcing my decisions to ChatGPT upped my empathy for all of us struggling to make good choices in a culture that is always pushing us, with everything from corporate mindfulness retreats to the latest diet trends, to be fitter, happier, and more productive. It’s hard enough just being human, what with all the uncertainty—and, okay, existential dread—built into the gift of free will. So for now, at least, I’ll just try to enjoy my pizza and be glad the robot overlords aren’t programmed to order us around (…yet?). And when I’m craving the input of a judgy AI, there’s always Co-Star.

— Ali Shapiro