Learning from GPS: How Not to Lose Our Way with AI

Recently I was in Connecticut visiting family, relying on GPS to get everywhere. Like most of us, I’d type in the destination and pick the route with the shortest ETA.

Decision made, thinking outsourced.

One night after dinner, I followed my cousin back to his house. I had the GPS running, but instead of following its directions, I followed him. Turn after turn, the GPS kept trying to “correct” me—redirecting, recalculating, insisting I was off course.

We ended up taking a route that wasn’t one of its original options. When we pulled in, I checked the time. We arrived a minute faster than any of the GPS suggestions.

That moment stuck with me. Not because we “beat” the algorithm, but because it made me realize how rarely we even give ourselves the chance to try. Researchers have noticed the same pattern: GPS automates decisions our brains used to make, which can slowly change how we build mental maps of the world (Fenech et al., 2021).

What we lose when we stop getting (a little) lost. GPS gives us routes. We choose the fastest one. And then we follow it—often without question. But what have we given up in that exchange?

“With GPS, we still arrive at our destination—but we often lose the mental map of how we got there.”

Several studies now show that when people navigate with GPS or similar guided systems, they form weaker spatial memories and less accurate mental representations of their environments than when they find their own way or use maps (Gardony et al., 2018; Dahmani & Bohbot, 2020). In one experiment, the more time participants spent looking at a GPS-like map while learning a virtual city, the worse they were later at finding their way without it and at recalling landmark locations (Gardony et al., 2018). Another study found that habitual GPS users had poorer spatial memory when they navigated on their own, even though they could still follow routes (Dahmani & Bohbot, 2020).

A recent systematic review and meta-analysis went further, showing that frequent GPS use is associated with slightly worse “environmental knowledge” and a weaker sense of direction, even if wayfinding performance—the ability to reach a destination—doesn’t suffer (Javora et al., 2024). In other words: we still arrive, but we don’t really learn the place.

That’s exactly how my GPS trip in Connecticut felt. I got everywhere efficiently. But when I followed my cousin, ignoring the “redirecting…” prompts, I suddenly had to pay attention—landmarks, turns, patterns. The route became something my brain had to own, not just execute.

I felt this again the other night driving home from my best friend’s place—a route I’ve driven hundreds, if not thousands, of times. I usually turn on GPS just to check traffic. This time, I turned it off.

It felt… freeing. And allowed my brain to reflect.

What if traffic popped up? Would I reroute on my own? Would the hassle be worth the two minutes I might lose? Even asking that question felt unfamiliar. Because increasingly, we don’t even make that decision. We default to the machine.

Many of us remember a different experience. Printing directions on MapQuest before the trip to the mall. Sitting at AAA with our parents while they mapped out the route and handed us a stack of highlighted maps. Those weren’t just logistics. They were training. They gave us the entry-level skills we’d need when we were eventually handed the keys.

And maybe the best part of traveling wasn’t the efficiency—it was getting lost. Discovering a café you never planned to visit. Finding your way back using intuition, clues, and a bit of trial and error. Those moments weren’t mistakes. They were the point.

Cognitive Offloading: From Roads to Ideas

Psychologists have a name for what GPS is doing for us: cognitive offloading. We move memory and decision-making from our minds to external tools. With navigation, that means the device holds the map, chooses the route, and adjusts when things change. Our brains no longer have to build and update a rich “cognitive map” of the environment.

Now we’re seeing the same thing in a different domain: thinking.

Generative AI tools—ChatGPT, Gemini, Copilot—are becoming the new GPS for writing, problem solving, and creativity. They suggest outlines, generate drafts, debug code, and answer complex questions in seconds. They are astonishingly useful.

And they create the same temptation: why struggle through a messy first draft or a hard problem when a tool can give you a polished route to the answer?

“Moderate collaboration with AI tends to boost creativity; heavy reliance on AI tends to flatten it.”

Studies in education and learning are starting to map out this trade-off. One line of research on AI in classrooms describes a “cognitive paradox”: AI can scaffold critical thinking when used deliberately, but when students use it mainly for quick answers, it shifts learning from constructing knowledge to consuming answers (Vijayakumar et al., 2025; Surviyana, 2026). A recent qualitative review of AI in learning found that unregulated use often substitutes students’ reasoning rather than supporting it—reducing analytical engagement and weakening processes like evaluation and reflection (Surviyana, 2026).

Put simply: if the tool does the hard thinking, the student gets the grade, but not the skill.

Other work looks at creativity and problem solving. Experimental studies with workers and professionals show that AI can boost creative performance—but only up to a point. When people collaborate moderately with AI (rather than not at all or almost entirely), they generate more diverse ideas and produce more creative solutions; when they rely too heavily on AI, creativity drops and ideas converge (Gino, 2025). Similar emerging research with students finds that while AI can help them generate more concepts, it often pushes them toward similar patterns and can undermine their own creative confidence if it becomes the default source of ideas (Zhao & Zhang, 2025; Mou & colleagues, 2026; Agnaou, 2025).

It’s GPS all over again.

We still get to the destination—the report is written, the design is made, the assignment is submitted—but we may not be building the deep skills underneath: structuring arguments, navigating ambiguity, and discovering original routes through problems.

The Lost “Permit Phase” of Learning

In the workplace, especially in learning and development, we’ve traditionally relied on entry-level roles to provide exactly that kind of experience. Safe environments to struggle. To try. To fail. To build judgment.

A kind of “permit phase” before being handed the keys.

Those early wrong turns—messy drafts, clumsy stakeholder emails, imperfect analyses—are the cognitive equivalent of getting a little lost in a new city. They let people build mental maps of the work: how decisions are made, which paths lead to dead ends, where value actually gets created.

“If we skip the ‘permit phase’ of learning and go straight to automation, we risk creating professionals who can operate the tools but can’t navigate without them.”

When AI tools enter too early and too strongly into that phase, something subtle but important can happen. If a new analyst, instructional designer, or HR partner always starts with an AI draft, they may ship competent work without ever building their own internal repertoire of patterns and judgment. Some recent studies of students using AI for problem-solving show exactly this: they complete tasks more quickly, but display shallower understanding and weaker independent problem-solving when the tools are removed (Vijayakumar et al., 2025; Surviyana, 2026; Agnaou, 2025).

It’s the equivalent of going straight to autonomous driving without ever learning how to drive. That’s not a technology problem. It’s a design problem.

Designing a Healthy Relationship with GPS and AI

For those of us in L&D, talent, and people leadership, the question isn’t whether to use these tools. Of course we should. The evidence is clear that both GPS and AI increase efficiency and can, when used well, free up mental resources for more complex tasks (Fenech et al., 2021; Gino, 2025).

The question is: how do we build a healthy relationship with them? What skills are we comfortable offloading entirely? Which ones do we need to own? And in the middle—where do we need people to build real competence before introducing the tool as an accelerator?

Research on navigation suggests a few useful principles. People who use GPS selectively—leaning on it in unfamiliar environments but engaging more actively in familiar ones—end up with better spatial knowledge than those who treat it as a universal autopilot (Fenech et al., 2021; Javora et al., 2024). Similarly, studies on creativity show that moderate collaboration with AI, rather than full outsourcing, yields the best creative outcomes (Gino, 2025).

“The real risk of GPS and AI isn’t that they make us less capable; it’s that they quietly remove the small challenges where capability is built.”

We can bring that logic into our learning and talent strategies:

• Build foundations first. Design roles, curricula, and onboarding so that people have to wrestle with problems themselves before AI becomes a standard part of their workflow (Surviyana, 2026; IntechOpen, 2026; Agnaou, 2025).

• Make AI a partner, not a driver. Encourage “moderate collaboration”—use AI for brainstorming, alternative perspectives, or surface-level drafting, but keep core reasoning, structuring, and final judgment firmly in human hands (Gino, 2025; IntechOpen, 2026; Agnaou, 2025).

• Protect practice time. Just as getting a bit lost in a city helps us learn it, we need protected spaces at work where people can think, write, and solve without immediately reaching for AI, especially early in their careers (Vijayakumar et al., 2025; IntechOpen, 2026).

• Teach meta-skills. Help people understand when a tool is appropriate, and what they’re deliberately choosing to offload, so they can maintain ownership of higher-order skills like critical thinking, ethical judgment, and complex problem solving (Surviyana, 2026; IntechOpen, 2026).

Because context matters.

Drop me in the middle of a desert without GPS, and I might eventually figure out direction—but I’m still in trouble.

Drop me in New York City—a place I spent years navigating before smartphones—and I’d be just fine. I’d use landmarks, patterns, memory. I’d get to my Broadway show, and probably still have time to grab a slice before it begins.

That confidence didn’t come from a tool. It came from experience.

And that’s the part I don’t want us to lose. Not because efficiency is bad. Not because technology is the enemy. But because there’s something important in the process of finding your own way—on a map, in a career, in a hard problem.

Even if it takes a little longer. Even if you get it wrong a few times. Even if the GPS—or the AI—keeps telling you to turn around.

Sometimes, the best route is the one you learn how to navigate yourself.

Aspect	GPS / Navigation tools	AI tools for thinking
What they offload	Choosing routes and building a mental map of physical space.	Drafting, summarizing, idea generation, and problem-solving.
Main upside	Faster, easier travel, especially in unfamiliar places.	Faster writing and analysis; more ideas; support for novices.
Main Risk	Weaker spatial memory and sense of direction when used all the time.	Shallower understanding and reduced independent critical thinking with heavy reliance.
Lost learning moments	Getting a bit lost, improvising, and learning a city by feel.	Wrestling with first drafts, building arguments, and exploring unconventional solutions.
Healthy use pattern	Use often in new places; sometimes turn it off in familiar ones to keep your “navigation muscles” active.	Use for support and brainstorming; keep core reasoning and final judgment human-led.
What L&D should protect	Ability to navigate without constant turn-by-turn guidance.	Ability to think, write, and solve without defaulting to AI for every step.

Written by Erin Miller with support from Perplexity (AI assistant)

---

References:

GPS, navigation, and cognitive offloading

• Fenech, E., Wiener, J. M., & Harris, M. A. (2021). Rethinking GPS navigation: Creating cognitive maps through navigation. Frontiers in Human Neuroscience.

• Gardony, A. L., Brunyé, T. T., Mahoney, C. R., & Taylor, H. A. (2018). Spatial knowledge impairment after GPS-guided navigation: Eye-tracking study in a virtual town. Spatial Cognition & Computation.

• Dahmani, L., & Bohbot, V. D. (2020). Habitual use of GPS negatively impacts spatial memory during self-guided navigation. Scientific Reports.

• Javora, V. et al. (2024). GPS use and navigation ability: A systematic review and meta-analysis. Current Research in Behavioral Sciences.

AI, creativity, and critical thinking

• Vijayakumar, S. et al. (2025). The cognitive paradox of AI in education: Between enhancement and erosion of critical thinking. Frontiers in Education.

• Surviyana, S. (2026). The paradox of artificial intelligence in learning. Proceedings of Educational Technology and Development.

• Gino, F. (2025). Unlocking creativity with artificial intelligence (AI). Journal of Applied Psychology.

• Zhao, Y., & Zhang, L. (2025). The mediating role of creativity self-efficacy in generative AI use and college students’ creative problem-solving. Open Journal of Social Sciences.

• Mou, T. Y. et al. (2026). An exploratory study of students’ experiences with AI tools in visual design tasks. Thinking Skills and Creativity.

• Agnaou, A. (2025). Impacts of large language models on creativity, critical thinking, and problem-solving. Journal of Educational Research & Practice.

• IntechOpen chapter (2026). The pedagogical paradox of AI: Sustaining critical thinking through activity-based learning. In: AI and Education.