A suggestion box that rewards useful criticism

Businesses, public agencies, organizations, and individuals can miss easy improvements that their customers, clients, citizens, or colleagues would value. Someone may have spotted a change that could save thousands of pounds, or just make everyone's day easier. But telling the organization isn't always worth the effort.

That's understandable. Useful criticism takes time to write. It can make people angry, and sometimes the critic gets punished for it. It's rarely rewarded. I've been calling constructive suggestions of this kind complainments: part complaint, part compliment.

Why rewards are hard to promise

An organization might want to pay for good suggestions, but it can't easily make that promise credible. Once it has a valuable idea in hand, it has a reason to decide the idea wasn't worth much and pay little. Contributors can see this coming.

Running the box is also work: sorting submissions, clarifying vague ones, judging which are useful, implementing them, reporting back, and paying people. An ordinary anonymous form may also leave no way to reach the contributor with a question or a reward.

The proposal

A third party runs the box. It keeps contributors' identities and contact details from the organization, keeps a way to reach them, and administers a reward budget the organization committed in advance.

1. Precommit funds. Before any suggestions arrive, the organization commits or pre-funds a monthly or yearly reward pool, to be paid out under published rules.
2. Submit protected feedback. Contributors send suggestions through the administrator, which removes identifying details before passing them on. The wording itself can still offer clues, so contributors need control over what is shared.
3. Independent review. The administrator assesses usefulness and implementation, or requires the organization to report on and score each suggestion.
4. Pay and report. Rewards come out of the committed pool under the rules, and the administrator reports what was paid and why.

The commitment is the important part. If the money has already been set aside and must be paid under published rules, rather than quietly reclaimed, being stingy after the fact gets harder. A deposit held by the administrator could support this. The pool would guarantee an overall commitment, not a particular payment to everyone who submits or a precise share of the value their idea creates.

The administrator's fee should be disclosed and kept separate from the money committed to contributors. Rejected claims shouldn't send money to the administrator, or it would have a reason to reject them.

This won't solve everything. An administrator can't verify every claimed benefit, and it can't rule out collusion between an organization and a friendly contributor. Oversight and incentives need careful design. Several rules are still open: what happens when there are few or no qualifying suggestions, where unused funds go, and how to handle duplicates, disputes, and valuation.

Where AI might help

Language models could make this less work: sorting submissions, grouping duplicates, helping contributors clarify their feedback (with their approval), and helping audit how the process works. The aim would be to make decisions inspectable while protecting contributors' identities. Models can invent details or lose nuance, so people would make final payout and dispute decisions. Comments sent through this page aren't automatically shared with AI tools.

A possible pilot

Many organizations and individuals in effective altruism already ask for anonymous feedback. Are people actually using these boxes, and are the incentives enough? The Unjournal, which I co-direct, could volunteer as a test subject, and might help administer pilots for others. For feedback about The Unjournal itself, an independent outsider would need to administer the rewards. This is a possibility, with no commitment from The Unjournal yet.

What I'd like to know

• Has something like this been tried? What happened?
• What reward rules would you find credible, as a contributor or as an organization?
• How would it fail? Gaming, collusion, too few submissions, legal or tax problems, something else?
• Would you use it, test it, fund a pilot, or help design or run one?