Why group chats are terrible at making decisions

You know the shape of it. Someone suggests dinner. Four people send an enthusiastic reply within a minute. Somebody proposes a restaurant, someone else proposes a different one, three people react with a thumbs up to something ambiguous, one person asks “which weekend are we talking?”, and the thread goes quiet. A week later somebody posts “so are we doing this or…” and the cycle restarts from the top.

The usual diagnosis is that people are flaky. That is almost never what is happening. Group chats fail at decisions for structural reasons, and the same eight people will happily close the same decision in ten minutes if you change the container it happens in.

A chat cannot tell a no from a not-yet

In a ballot, silence has exactly one meaning: this person has not voted. In a chat, silence has at least four. It might mean no. It might mean not yet, I am at work. It might mean I never saw it, because the thread had ninety unread messages. It might mean I assume this does not need me specifically.

Because the organizer cannot distinguish between them, they do the only safe thing and wait. And waiting is exactly wrong for three of the four cases. Most stalled plans are not waiting on a decision; they are waiting on a signal that will never arrive because the person who owed it does not know they owe it.

The options never stop moving

A vote requires a fixed set of choices. Chats refuse to fix anything. Message 12 proposes three restaurants; message 31 adds a fourth; message 44 revives the second one with a different date attached. Nobody in the thread is evaluating the same question, because the question changed between when they read it and when they replied.

This is why chat threads produce so many stale agreements — someone says “yes!” to a plan that has since mutated, and their yes gets counted for a version of the evening they never actually agreed to. The organizer ends up holding a pile of approvals for slightly different plans and no way to add them up.

There is no closing time

Chats have no end state. They have a most recent message. Nothing in the medium ever announces this is decided, so decisions get made by exhaustion instead: whoever is most persistent, or whoever books something unilaterally and presents it as settled, wins.

Deciding by exhaustion has a cost people underrate. The person who ends up booking is usually the same person every time, and the reason they eventually stop organizing things is not that the plans were bad. It is that closing a decision in a chat requires being slightly annoying in public, repeatedly, and nobody wants that role permanently.

Nobody in particular is being asked

A message to twelve people is a message to nobody. This is the bystander effect in its most ordinary form: the larger the group, the lower each individual’s sense that the reply has to come from them specifically. Response rates fall as group size rises, and they fall faster than most organizers expect — the same question that gets four replies out of five in a group of five gets six out of fifteen in a group of fifteen.

Addressing people individually fixes the psychology but breaks the logistics: now you are running twelve private conversations and reconciling them by hand. What you want is a message that is addressed to one person and counted centrally, which is precisely what a chat cannot do and a poll invitation does by default.

Early agreement suppresses honest disagreement

Chats are public and sequential, which means the first two replies shape everything after them. If two people say “love it” to the expensive restaurant, the third person — who cannot really afford it this month — is now choosing between an awkward objection in front of everyone and a cheap thumbs up. Most people pick the thumbs up.

You end up with visible consensus that is not real, and it collapses later in the form of last-minute dropouts. The dropouts look like flakiness. They are actually the delayed expression of an objection the thread made too expensive to state at the time.

A private ballot removes the audience. People answer what is true for them rather than what is socially cheapest, and you find out about the problem while it is still a scheduling question rather than a cancellation.

Saying no costs too much

Related, and worth separating out: in a chat there is no graceful, low-cost decline. Writing “I’m out” into a thread full of enthusiasm feels like puncturing something, and it invites follow-up questions you may not want to answer. So people go quiet, which is socially free and informationally useless.

Every group needs declining to be a first-class action — one tap, no explanation required, no audience. The moment that exists, the ambiguous silences mostly disappear, because the people who were staying quiet to avoid a scene now have something easier to do than nothing.

What group chats are genuinely good at

This is not an argument for abandoning the chat. Chats are excellent at the two phases that sit on either side of the decision:

• Divergence. Generating ideas, riffing, discovering that three people have wanted to try the same place. A poll cannot do this; you need the messy conversation to find out what the options even are.
• Logistics after the fact.Once the decision exists, chats are the right place for “running ten late”, “table is under my name”, and the photos afterwards.

The failure is using one medium for all three phases. Divergence and convergence want opposite properties: open versus fixed, conversational versus countable, ongoing versus closing.

A protocol that works

1. One person owns it.Say so out loud — “I’ll run this”. Ownership is what converts a topic into a task.
2. Timebox the ideas.“Throw suggestions in until tomorrow night, then I’ll put the top few to a vote.” The chat now has a job and an end.
3. Freeze three to five options. Write them down concretely, including dates and times. Late suggestions go in the next round, not this one.
4. Ballot, privately, with a deadline. Everyone gets the same fixed question, answers without an audience, and can decline in one tap. State what happens when the clock runs out.
5. Announce the result in the chat, with the count.“Thursday, seven of nine. Booked.” The number does the work — it makes the outcome look like arithmetic instead of a preference, which it is.

The underlying point

Every one of these failures is about the shape of the container, not the character of the people in it. A thread that never freezes a question, never closes, never distinguishes silence from refusal, and charges social capital for disagreement will produce indecision no matter who is in it.

Give the same group a fixed question, a private answer, a visible decline, and a deadline, and the flakiness mostly evaporates. It was never really flakiness. It was a chat being asked to do a job it was never built for.