When a Voice Guides Our Choice

Imagine you’re at a restaurant. The server rattles off the specials, and by the time they finish, you’re already struggling to recall the first dish. Now imagine those same specials printed on a menu. You can glance back, compare ingredients, and weigh your options at your own pace. While the difference may seem trivial, new research shows it can shape not only how we evaluate information, but also the choices we make.

With the rapid rise of voice assistants like Alexa and Siri, more of our decisions are being mediated by spoken words rather than written text. From shopping to customer service, we’re increasingly listening rather than reading. In a new working paper, I investigated this topic with my co-author Vicki Morwitz from Columbia University. Our studies, involving more than 4,000 participants, reveal that this shift in medium has subtle but meaningful effects on judgment and decision-making.

The reason lies in how our minds process information. Text has an advantage: it stays put. You can reread a sentence, compare numbers side by side and look back if you forget. Speech, in contrast, is fleeting. Once words are spoken, they vanish, leaving listeners to hold the details in memory. This difference becomes especially important when we need to compare options, a situation where memory matters most.

For comparative judgments, some information is easy to interpret on its own, like when you see “brand new” or “recommended” in a product description. Other information only makes sense with comparison. Are 20,000 entries in a dictionary a lot? Only if you know that another dictionary has 10,000. This latter kind of information, known as low-evaluability information, plays a central role in many real-world choices.

Our experiments show that listeners struggle more than readers to use low-evaluability information when comparing options. As a result, their evaluations tend to be less extreme. When the information is favorable, listeners are less enthusiastic than readers; when it’s unfavorable, they are less harsh. In one study, for example, participants were told about a multivitamin’s ingredients and whether they met or missed recommended benchmarks. Readers were quick to spot when the vitamin was better or worse than the benchmark. Listeners, however, found it harder to keep the details straight, and their ratings fell closer to the middle.

This difficulty doesn’t just affect opinions, it also influences behavior. In another experiment, participants were asked to reorder snack bars described in either text or voice. The previous product was unavailable, so a new one was offered that was clearly an upgrade: larger bars, better nutrition, and higher customer reviews. Most readers chose to accept the substitute. But listeners were significantly less likely to do so, apparently because they struggled to process the comparative advantages.

The implications extend to consumer behavior and beyond. For companies betting on voice shopping, our findings suggest a hidden barrier: customers may be less likely to switch to better products if they only hear about them. But there’s also a potential solution. In one follow-up test, we found that phrasing offers in explicitly comparative terms (“this product has 10 more grams of protein than your previous choice”) helped listeners bridge the memory gap.

The results also highlight how context shapes decision-making. Evaluability theory predicts that people often reverse their preferences depending on whether they evaluate several options at once or one at a time. Our work shows that these classic “joint vs. separate” preference shifts are less pronounced when people listen instead of read, again because listeners can’t hold all the comparative details in mind as easily.

Listening, of course, also has strengths. A human voice can carry warmth, urgency or persuasion in ways that text cannot. That may be why people are often more influenced by recommendations delivered out loud. But when it comes to carefully weighing trade-offs, speech can make it harder to use important (but less easy-to-evaluate) pieces of information.

As voice technologies continue to spread into shopping, healthcare, education and everyday decision-making, it’s worth remembering that listening and reading are not interchangeable. Each mode nudges us toward different ways of thinking. Our findings suggest that designers of voice systems should take these differences seriously. Sometimes, the smartest choice may simply be to put it in writing.