Thank you, Jason! Until we can find a mitigation, we'll update our website to inform donors about the problems you found. We appreciate your thoughtful contributions :)

Excellent idea, we'll set our creative talent on it. If the result isn't too amateurish we'll include it in our About page. 

We hadn't considered this framing! DuelGood wouldn't contribute anything if everyone who uses it would have donated to GiveWell anyway. And if Person A was already planning to donate to GiveWell and uses DuelGood to also block a political donation, Person A hasn't made a real sacrifice while their matched partner has. This asymmetry could undermine trust and discourage participation.

I wonder if the net outcome might still be positive. 

• Consider Person A, who was going to donate $100 to GiveWell anyway and now donates through DuelGood to neutralize a gun safety donation.
• Person A's $100 still go to GiveWell, same as if they had donated directly.
• In addition, a gun safety donation is blocked and redirected to GiveWell instead. This is a net positive assuming A: GiveWell is more effective than donating to gun safety, and B: perceived gaming doesn't reduce participation enough to eliminate the benefit.

The system creates value as long as someone on each matched pair is making a counterfactual change. The question is whether perceived gaming reduces participation enough to outweigh these benefits. 

We don't have a perfect solution to the problem of trust, but here's some initial things I can think of that might help:

• Include testimonials from previous donors making sacrifices.
• When donating, allow people to specify if they had already heard of GiveWell/select from a list like "What best describes your donation: (a) Redirecting a planned political donation, (b) New donation inspired by DuelGood, (c) Already planned GiveWell donation". This would help us gather data on how many people are potentially gaming and then update.
• Include this problem on our FAQ to improve transparency. 

This may be a limitation we can't fully solve. If enough people are gaming, the concept may not be viable at scale. We'd value your thoughts on mechanisms that could help.

(p.s. thank you for your input!)