When Memory Leads to Wrong Moves

“If the present looks a lot like the past, then of course you’d want to rely on what you learned in the past to make the best possible decision,” says Jörg Spenkuch, an associate professor of managerial economics and decision sciences at the Kellogg School.

“But what happens when the environment changes, when we find ourselves in a new or unfamiliar situation?” asks Yuval Salant, a professor of managerial economics and decision sciences, also at Kellogg.

Salant, Spenkuch, and PhD student David Almog explored this question in the context of chess, a game where people rely heavily on their memory to make high-stakes, time-sensitive decisions.

They specifically examined how memory influenced players’ opening moves the first time they played Chess960, a variant of traditional chess where the starting positions of pieces in the back row are scrambled. When playing this altered version of the game, players chose to use an opening move from their memory the vast majority of the time—even if it was not the best move for the situation at hand.

The findings suggest that, when we have to make a decision in unfamiliar situations, our instinct to rely on memory generally interferes with our desire to make the best possible decision. The economists refer to this phenomenon as the “memory premium.”

“Decision-makers tend to choose options that are retrievable from memory over those that are not,” Spenkuch says, “even when the decisions that are not in our memory are, in principle, better.”

A sweet spot

For their research, the economists partnered with the popular internet chess server Lichess, which hosts millions of standard chess games and thousands of Chess960 games every day.

They obtained information on all moves made on the server by individual players from January 2013 through June 2021. And they zeroed in on the opening moves of nearly 150,000 first-time Chess960 players with a history of playing traditional chess. Moves that an individual player had used at least once before in traditional chess were considered to be part of their memory.

Chess960 involves the same board, pieces, and rules as traditional chess but randomizes the initial positions of pieces in the back row, making it very impractical to memorize and deploy opening sequences the way many players do for standard chess. Without the heavy advantage of memorizing moves, players are encouraged to analyze the situation in real time instead.

What’s more, Chess960 allowed the researchers to objectively measure the quality of players’ decisions using sophisticated computer algorithms. This addressed concerns that players might have chosen moves from their memory simply because they were better than the ones that weren’t in their memory.

Taken together, these features gave the team an ideal setting to investigate the influence of memory on decision-making.

“We think of chess as the sweet spot between a tightly controlled but somewhat abstract, unrealistic lab experiment and decision-making in the wild, where we have to impose all sorts of heroic assumptions about what decision-makers know and are trying to do,” Spenkuch says. “It’s a controlled environment, but also real-world data.”

The memory premium

After analyzing their large dataset, the economists found that people who played Chess960 for the first time used opening moves from their memory roughly twice as often as moves that they had not played before. This occurred even in cases where the memory-based moves proved to be objectively worse.

Salant, Spenkuch, and Almog also confirmed that the memory premium in this chess environment mirrored many of the patterns that psychologists and neuroscientists associate with memory retrieval in the real world.

The features of this behavior “are hard to explain with anything other than memory,” Salant says.

The Chess960 players, for instance, were more likely to use moves that they had used more recently and more frequently while playing standard chess—reflecting the easier time people have recalling fresh or recurring experiences. The memory premium was also stronger when the placement of pieces on a Chess960 board more closely resembled the arrangement in standard chess.

In addition, newer memories seemed to disrupt the recall of older memories. As people played more games of Chess960, they became less reliant on their pool of moves from standard chess to make their first move. After about 25 games of Chess960, the memory premium decreased by nearly 50 percent.

“Good enough”

Collectively, the findings underscore the fundamental role of memory in how we make choices.

But why do we so often turn to choices from our past even when they are not ideal for our current situation? We simply might not know a better way to respond to the unfamiliar situation, and relying on memory is generally easier than finding the best possible decision in real time.

“To find the potentially best starting moves [in Chess960] would take time and effort, and it’s not clear that even if decision-makers tried that, they would be successful every time,” Spenkuch says. “So drawing on memory, even in unfamiliar environments, may well be rational in the sense that it economizes on effort and costs—a kind of cognitive shortcut.”

Another reason may be that we tend to make decisions based on the first choice that comes to mind, which, in many circumstances, might come from our memory.

“So, if I see a new situation, and an option comes to mind, I ask myself, ‘Is this good or not?’” Spenkuch says. “If it looks like a good enough choice, I take it. I don’t bother thinking about more options.”

Fortunately for decision-makers, the researchers found that players who chose moves from their memory actually tended to perform a little bit better than those who didn’t—except in the most unfamiliar situations.

“Those choices [from memory] are still pretty far away from the theoretical optimum,” Spenkuch says, “but finding the best move every single time, that’s probably too much to ask of humans.”

Avoiding mishaps

Being aware of these tendencies can help us make more-informed decisions and avoid mishaps when we encounter new situations, the researchers say. Uber’s global expansion offers a case in point.

The ride-sharing technology company found great success in many U.S. cities by entering the market without legal permission, ballooning its audience base in that region, and then, over time, cajoling local officials into negotiating agreements for the company to stay there.

So when Uber made the call to expand to Germany, it decided to take the same approach that had worked for them so well before. It started offering its services in various cities, flouting laws and regulations, and then contacted local officials to win their support. But this time, the strategy flopped. Uber was sued and sentenced to outrageous fines. Local officials, it turns out, had very little leverage in Germany and were unable to provide the company with the protections it needed to grow. So Uber had to leave these cities until it could hammer out the appropriate legal framework.

“When you extrapolate market entry in one environment to market entry in a different culture or geography, if you only rely on what you did in the past, that may backfire on you,” Salant says. “Not all markets are created equal or similar.”

That’s not to say that when we’re faced with a tough decision, we should entirely forgo our past experiences. Uber did find success using the same strategy in other markets, as did many of the chess players who used their old opening moves for Chess960. But it’s important to recognize that different situations may require different choices.

“Memory retrieval is something that should work on average,” Salant says. “The surprise is that we keep using this heuristic even when the situation is very different.”