Inspiration

Voting is hard. No one feels good casting a ballot not knowing half the names, but at the same time it's a daunting amount of work to actually research everyone fairly. Enter Wishing Well. We strive to make the process of researching and deciding on what you care about easier. We make it simple for people to look over what is on their ballot and gameify the process of making your choices so that by the end, you're confident and ready to make your wishes heard.

Scope

Automatically process every contest on the Travis County, Nov 3 2026 ballot (early voting Oct 19-30). The pipeline attempts full evidence treatment for each; only a contest that fails source/quality gates falls back to objective information. Architecture supports other jurisdictions via adapters. We host an Expo + React Native mobile app, plus a small web page for share links.

A big part of the app is hiding the candidates' party and personal identity, and only showing their advertised platform and prior policy actions to compare what they say to what they do, while taking out any personal biases. This is an extremely important psychological bias to remove, which can't really be done in any other form of political information gathering.

Challenges

The biggest challenge we had wasn't actually in our scraping or data gathering (though those were plenty challenging) so much as considering the psychology of how users engage with politics. We want our users to be able to easily pick up and use our app; the mental friction should be low. However, this induces a feeling of cognitive ease, which can be particularly emphasized if they feel an overload of information. This then leads them to be more susceptible to cognitive biases such as picking "Cinderella" over "The Big Bad Wolf". Things like how we phrase a question or what options we give the user have psychological impact on people's tendencies towards certain decisions, and we need to take those into account when building the design out.