Attention Is Not Demand

Promising is not the same as marrying.

— Russian proverb

On one screen, 327 likes; in the calendar, not a single date. The contradiction exists only for as long as a like is treated as a scaled-down copy of intent. But a like is not a small proposal of a relationship. It is a separate action with its own price, usually a very low one: marking a profile takes a second, replying takes a minute, keeping a conversation going costs more, and showing up for a date means giving up an evening, travelling there, sitting through the awkwardness and dropping other plans. A second date costs more still, because a continuation now competes with other options and has consequences. The flow of attention and the number of available partners diverge for a reason; they measure different steps of the same movement.

The error shows up nicely in an advertisement for a flight, “Paris to London for 49 euros*”. The asterisk led to a long footnote, but the phones started ringing immediately. Marketing could report a rise in enquiries, the call centre could report being swamped, and the flight was not necessarily filling up with passengers ready to buy a real ticket on real terms. The campaign increased attention, but not always demand. Dating apps let you repeat the same error in your personal life, except that instead of calls the board shows matches, profile views and the glow of notifications.

Five doors after the like

Imagine a thousand profile views. Some of those viewers will mark it; of the matches only some will start a conversation; of the conversations some will live longer than three messages; a few will lead to a date; and fewer still to a mutual wish to continue. A new constraint operates at every transition. Before the like, what matters is the photo, the first lines and the position in the feed; before the message, mutual interest and some spare attention; before the date, goals, schedules, distance and basic trust; before the second date, the experience of that particular pair. A large flow at the top says nothing about where exactly the funnel narrows. It can coexist with poor-quality matches, an unclear signal about goals, or simply with the platform showing the profile to people who mark willingly and reply rarely.

In signalling theory an action is informative when it is not equally easy for people of different types to perform. A degree conveys both the content of a course and the ability to complete a long one; a manufacturer’s warranty carries information if it costs a bad manufacturer more because of future returns. A like costs almost nothing, so it can be sent by someone with serious interest, by a bored commuter and by an automated account alike. A detailed reply to a specific detail of a profile requires attention; an invitation to meet requires intent; arriving on time takes a little more effort still; and repeated behaviour gradually turns separate signals into a reputation. Time, specificity and a willingness to give up some alternatives raise the informativeness of an action, though they guarantee no good intentions.

Same button, different goals

Even an expensive step is hard to interpret without knowing the goal. One person is looking for marriage, another for sex, a third for confirmation of his own attractiveness, a fourth has not decided yet; all of them can press the same button and invite you for the same coffee. The platform collapses different contexts into a single interface. In an ordinary social setting some intentions are visible in advance: meeting at a professional conference, in a nightclub and through relatives all start from different expectations. In an app the context has to be reconstructed through words and a sequence of actions. That is where the belief comes from that “everyone only wants one thing” or, conversely, that “nobody knows what they want”: a participant sees a stream of identical signs standing for different aims, and mistakes the ambiguity of the channel for sameness in people.

An excess of attention also has a limit to its usefulness. When there are more messages than time, replies get shorter, selection gets cruder, and each correspondent gets fewer chances to show anything. The recipient starts using the cues available immediately: the photo, the profession, the wit of the first line. This is a rational response to overload and a poor way of seeing qualities that emerge slowly. Two people with very different like counts can therefore have almost the same number of real opportunities: one gets ten matches and talks properly with five, the other gets a thousand and manages a proper conversation with the same five. The top of the funnel differs by a factor of a hundred, the bottom hardly at all.

The message market sees competition, but not intent

Elizabeth Bruch and Mark Newman reconstructed the top of the funnel from messages in four American cities. Desirability was defined by the network: what mattered was the number of approaches and the standing of the senders themselves. Users typically wrote to people about a quarter above themselves on that scale, and the probability of a reply fell as the gap grew. The result was a map of how attention moves: participants reached upward, recipients filtered the flow.

The network scale described a profile’s position in a particular message market. It did not include the intent to build a long union, behaviour in a shared life, or how a specific pair reacted after meeting. Even a reply remained just the next transition in the funnel, not a scaled-down copy of a relationship.

A more carefully written message slightly raised the chance of a reply when reaching “upward”, but did not turn the market into a vending machine for correct phrasing. The main result lay elsewhere: where approaches are cheap and competition is high, a platform sees the direction of attention well. The user’s goal — mutual continuation of contact — sits several floors below.

Work on actual correspondence gives a more modest picture. The move to a first date is linked to a combination of features rather than to one right phrase: attention invested in the conversation, mutual self-disclosure, coordination and an explicit move to take the contact offline. These studies observe messages that have already been selected and do not establish a causal effect for each turn of phrase, but they do support the step model: a reply and a date are different outcomes, predicted by different features.

What to look at is where the losses occur, not the total number of signs of attention. Low visibility or an absence of marks points to a constraint at the entrance: the setting, the photos, the description, the algorithmic feed. Many matches with few conversations means the cheap signal is not turning into action. Conversations that never reach a date more often run into mismatched goals, distance or the absence of a next step. After the dates start happening, the profile recedes: information about actual interaction begins.

A like tells you a person made a cheap movement in your direction. That is better than no movement, and much less than intent. Attention becomes demand only when someone is ready to pay the next price — in time, in clarity and in giving up some alternatives. An advertising call is useful only as the start of a path to a purchase; a glow on the screen only as a chance to start getting to know someone.

Main sources

Spence, M. (1973). Job market signaling. Quarterly Journal of Economics, 87(3), 355–374.

Finkel, E. J., Eastwick, P. W., Karney, B. R., Reis, H. T., & Sprecher, S. (2012). Online dating: A critical analysis from the perspective of psychological science. Psychological Science in the Public Interest, 13(1), 3–66.

Bruch, E. E., & Newman, M. E. J. (2018). Aspirational pursuit of mates in online dating markets. Science Advances, 4(8), eaap9815.

Sharabi, L. L., & Dykstra-DeVette, T. A. (2019). From first email to first date: Strategies for initiating relationships in online dating. Journal of Social and Personal Relationships, 36(11–12), 3389–3407; Huang, S. A., & Hancock, J. T. (2022). Will you go on a date with me? Predicting first dates from linguistic traces in online dating messages. Journal of Language and Social Psychology, 41(4), 371–395.