Anusuya Datta
Mohamed Hassan via Pixabay
I watched Odysseus spend ten years trying to get home. In the ancient world, that used to be a believable story. Getting lost was ordinary, and sometimes it killed you.
We went from stars to compasses to satellites. Today, a phone can locate us, calculate a route, adjust it around traffic, and guide us through a city we have never seen. We solved getting lost so completely that we barely noticed.
It seems to have cost us something.
A 2020 McGill study of drivers found heavier lifetime GPS use went with poorer spatial memory when navigating unaided. A 2024 review of 23 studies linked GPS use to poorer knowledge of our surroundings and sense of direction—but not significantly to wayfinding. We still arrive. We just don’t necessarily know where we have been.
*
A route recommendation looks like information, but it can lead to serious consequences. In February 2025, Toronto City Council adopted a motion addressing congestion at the intersection of Eglinton Avenue West and the Allen Road Expressway. That congestion, it said, had caused “unprecedented traffic infiltration” on residential streets, as commuters hunted for shortcuts or were sent down them by Waze and Google Maps.
The year before, on one of those streets, city staff had advised that apps were not registering a rush-hour turn restriction. The city extended it to twelve hours a day, every day, in a form the apps would read. This despite Toronto sharing data with Google-owned Waze since 2017.
Then, in May this year, a Google Maps glitch sent drivers the wrong way up a one-way Toronto street, and the city put up a temporary “Do Not Enter” sign.
A physical sign to overrule the software.
Montreal heard the same complaint. In 2019, residents of one street nicknamed it “the highway” and suspected Waze and Google Maps of being behind increased traffic. The city had already signed a data-sharing deal with Waze in 2016.
These are not just isolated cases within Canada. By 2017, Fremont, California, had put commute-hour turn restrictions on its busiest residential shortcuts and shared them with Waze so the app would account for them.
According to Jane Macfarlane, director of UC Berkeley’s Smart Cities Research Center, the apps “don’t care whether the residential streets can absorb the traffic.” Cities feed the apps, measure themselves with navigation data, and still field the complaints.
Sometimes, maps are plain wrong. On British Columbia’s Mount Fromme, Google Maps showed a nonexistent trail, and, in 2023, North Shore Rescue pulled a stranded hiker off the cliffs there, the second such rescue in two months.
Google deleted the trail two days later.
These are errors. And errors get fixed. But they show something that holds even when the map is right. We go where it tells us—onto a quiet street, the wrong way up a one-way, out onto a cliff—without question.
*
Sometimes the change is deliberate. Since 2017, Parks Canada’s marketing has included a hidden gems campaign to spread visitors beyond the busiest destinations. By 2019–20, national park visitation was 4 percent above its 2016–17 baseline: up 1 percent at the seven mountain parks (Banff, Jasper, Yoho, Kootenay, Waterton Lakes, Mount Revelstoke, and Glacier) and 8 percent elsewhere. Parks Canada’s report calls that a campaign success.
Once, the traveller chose a destination, and the map helped get them there. Search blurred that distinction. Type “coffee near me,” and the map decides which cafés come first.
Generative artificial intelligence pushes further.
In March, Google launched Ask Maps, a conversational system that answers complex questions about places. It reached Canada in August. Google says it draws on information from more than 300 million places and tailors its answers using your past searches, the places you got directions to, and your saved lists.
A local-search consultant’s limited, early test of local-service searches in one area found Ask Maps generally returned three to eight businesses and spent its effort explaining why they fit. A screen full of pins becomes a shortlist. Something else has done the choosing.
The paper map asked us to interpret it. GPS learned where we were. Navigation software chose the route. Now AI can help choose the destination itself.
The question has moved from “How do I get there?” to “Where should I go?”
*
A restaurant with thousands of reviews produces abundant signals; a new place may not. Popularity can be evidence of quality—or simply of popularity.
A January study in the journal Machine Learning used real venue-visit data from New York and Tokyo to simulate what happens when recommendations repeatedly influence where people go. People following recommendations visited a wider range of places. Yet, in most of the systems tested, visits became more concentrated, because that new exploration mostly led to the same popular venues.
In the most extreme case, modelled on New York, the most popular tenth of venues drew 57 percent of visits to places people had not visited before, against a third in the original data. The algorithm can make the popular feel personal, even when we are sent to a restaurant we have never heard of. We don’t experience the recommendation as conformity.
It also gives the hunt for a “hidden gem” a contradiction: we ask the algorithm for somewhere other people haven’t found, and if it works often enough, the gem stops being hidden. The feedback loop is simple. A place gets recommended, people visit, the visits become data. The data feeds the next round of recommendations. But what happens to the places that never enter the loop?
*
None of this makes me want paper maps back. But Ask Maps is new to Canada, and there is no published research yet on what it does to where we go.
Every map leaves things out. Now the omissions can be personalized. And being on the map is no longer enough. A place also has to be legible to the system doing the choosing. Then it has to be chosen.
That changes what it means to be lost. For most of human history, we were lost when we could not find a place. Now a place can be lost because the system does not find it for us.
Odysseus spent ten years lost and still knew where home was. We can cross the country without a wrong turn. But we are less sure we could tell how we got there or what the map decided we did not need to see.
ANUSUYA DATTA is a journalist and editor based in Saskatoon, Saskatchewan. Her writing explores technology, geography, public policy, and the ways they shape everyday life. Anusuya currently works as a Content Strategist at EarthDaily. This article was written in her individual capacity, and the views expressed do not necessarily reflect those of EarthDaily or its affiliates.