Why Learning That a Partner Enjoys an Activity Can Bring It to an End

A shared activity can provide enjoyment and express care for another person. In Why Good Experience Can End Joint Activity, a November 2026 theoretical preprint, Weikang Zhang asks what happens when evidence of personal enjoyment weakens the meaning attached to participating for a partner. Under the model’s assumptions, learning can end an activity even though continuing would still benefit both partners on average.

This is a theoretical mechanism, not evidence that enjoyable experiences generally damage relationships. It depends on a particular assumed concern: joining becomes costly when interpreted as pursuing personal interest rather than responding to a partner.

When enjoyment changes agreement

The model has a proposer and a responder choosing between a new joint activity and an existing arrangement both accept. The responder sometimes directly prefers the new activity and sometimes prefers the existing one. Because they also value their partner’s benefit, they may willingly join when the activity is not their own first choice.

Joining carries an “attribution cost” that increases with the proposer’s belief that the responder personally wants the activity. Declining preserves the existing arrangement without that cost. Crucially, the model assumes that fully revealing personal interest makes joining too costly even when the responder directly prefers the activity.

Participation on unwanted days therefore supports participation on wanted days by making agreement less revealing. If unwanted-day acceptances disappear, joining only when interested reveals precisely the preference that makes participation costly. Agreement can unravel in both states.

Two different stopping points

Zhang distinguishes a participation threshold, above which no equilibrium of the current interaction sustains agreement, from a higher payoff threshold, above which participation across both preference states ceases to benefit the responder on average.

At baseline assumed payoffs, these thresholds are beliefs of 50% and 75% that the responder wants the activity. Between them, participation is infeasible despite retaining positive average value for both partners.

The averaging matters: accepting on an unwanted day above the participation threshold would impose a current loss. The result does not recommend overriding refusal. It describes voluntary choices eliminating an opportunity that remains valuable across changing preferences.

What the diaries contribute

A 21-day diary study of 121 couples informs the model’s preference variation and feedback accuracy. Participants reported sexual desire, judgments of their partner’s desire and sexual activity. A separate 30-day diary of 133 couples helps constrain preference variation.

The diaries do not measure the model’s proposals, attribution costs or payoffs. For the main feedback measure, judgments on days without sexual activity provide relatively weak evidence of the partner’s reported desire.

At baseline payoffs, this feedback satisfies a theoretical condition preventing experience from raising beliefs enough for stopping to protect the responder on average, throughout the diary-compatible persistence range at point estimates. Sampling uncertainty and alternative feedback constructions affect the guarantee near that range’s upper limits.

Why waiting may not help

The main repeated-interaction analysis prescribes equilibria of each day’s interaction while ignoring how today’s choices affect future learning and payoffs. This is a behavioral assumption, not an equilibrium for forward-looking partners. Section 9 finds profitable responder deviations when future consequences are counted while subsequent prescribed behavior remains fixed.

Under this assumption, with opportunities themselves conveying no preference information, stopping cuts off further public evidence. The belief about the responder’s enduring tendency to want the activity freezes. Daily forecasts approach that belief; if it remains at or above the participation threshold, waiting alone never restores feasibility.

In 36 diary-compatible simulated environments using this prescribed, current-payoff behavior, stops that waiting cannot end account for at least 95.1% of weighted blocked opportunities over 84 days. These are simulated outcomes, not observed rates among diary couples.

Under the same behavioral assumption, the main simulation finds that equally informative feedback arriving independently of participation reduces joint payoff loss from 14.8% to 8.4% of the no-report benchmark. Expected daily report counts are matched across the population. Reports can now reach stopped pairs, although outside feedback still leaves a loss relative to no reports.

Why more information is not always a remedy

A separate forward-looking model makes the responder’s enduring propensity public and assumes preferences persist longer than the diaries support. With exact participation feedback, someone may accept at a current loss to establish a lack of interest and lower future attribution costs. These acceptances also make joining less revealing of enjoyment, supporting participation on wanted days.

Free, exact daily outside revelation removes that incentive. At this model’s baseline, it lowers the best combined payoff within monotone Markov equilibria, even allowing arbitrary public information before acceptance under outside revelation.

With weak, diary-shaped participation feedback, exact daily outside revelation instead raises the best combined payoff within monotone Markov equilibria with local screening, a restriction governing how unexpected choices are interpreted. That improvement is not established for unrestricted Markov equilibria. Information can restore missing evidence or remove an incentive to participate; the conclusion depends on feedback strength, behavior and the equilibrium class.