As the USA moves to more mail-in voting -- and more efforts at voter suppression by strict enforcement of ballot spoilage rules rather than lax enforcement when the intent of the voter is clear, it seems that some apps and tools could do things to reduce that.

It's too late to do this for the current election at this point (though some of these things do exist) but Here are a few ways to address some of the problems likely to be caused by the major influx of mail-in ballots, particularly in the states which don't allow them to be processed until election day. One existing app is My Vote Counts which has some of this information. You can also do a web search for your state's official list or map of ballot drop locations and information. It will usually be an official site of the state government.

Fairly distribute poll-worker effort

You have so many election workers, at polling stations and at county offices who count mail-in ballots as well as personally cast ballots. Normally, because counting and summing in-person ballots is easier, with no need to verify signatures on each one, that often gets done first.

Instead, locations friendly to a more even count could set a rule allocating poll labour, and have them take turns doing each type of ballot to keep it even. So in the first round, count 10% of the mail-in ballots and 10% of the in-person. You don't proceed to the next batch of either until the work is done. This is true even if the "work" of counting in-person ballots is to read the numbers off an electronic voting machine.

Phone app to help voters make no mistakes

Summary: An AI-based open source tool for phone/PC which will examine most mail-in ballots and warn about any flaws which could cause the ballot to be rejected or not reflect the voter's wishes. No images are stored, except of the signed envelope. The app then guides the voter on correctly enclosing and sealing the ballot, confirms postage is correct and directs the voter to a good place to drop the ballot.

Ballot Checker would be an app for phones and PCs. You would scan your mail-in ballot with your phone or computer. It would analyze your ballot for you and detect any of the slightest hallmarks of a potentially spoiled ballot.

These would include markings outside the area of marking, poorly readable markings, markings for more than one candidate, and races with no mark detected. The tool would use neural networks trained to both segment and isolate the marking areas, and to look for all the things known to cause spoiled ballots under the most strict rules in any common voting district.

Voters would be given warnings about potential flaws in their markings, and get a read-out on screen of what their ballot says. This would include a list of races with no vote found (many of which will be intentional.

The app would play non-partisan videos, customized as much as possible to the location of the voter, showing how to correctly mark a sample ballot (with fake names and parties) and how to fold and stuff the ballot in the appropriate envelopes, and in particular, how and where to sign the ballot -- the largest cause of ballot rejections come from signing the wrong place or not getting envelopes right.

The app could possibly help check that you have signed in the right place. It could guide you for where to look for your signature of record, which may be the signature on your voter registration, or drivers' licence or otherwise. A really clever app, given a copy of that signature, might even be able to confirm you got it close enough.

Ballot mailing guide

The system would save the image of the signed outside of the ballot, in fact it would also hash the photo and get an authenticated timestamp (something blockchains are good for) to prove the photo was taken at that particular time. This could be useful in any dispute over ballot fraud, to link mailed in ballots reliably with voters, and confirm a received ballot came from the voter in question.

The app would tell the voter how much postage to put on their ballot. After stamps are applied, it would confirm the postage is adequate.

The app would then show the voter nearby mail drop locations and ballot drop locations, and guide them to the locations with phone navigation. This would help voters avoid any false ballot drop locations.

All tools would be open source. Strict oversight would assure the app never makes any partisan advice, and only politely mentions races not voted in. To simplify the detection of races not voted in, the app could also establish an "official" way to mark such a race in a way that does not spoil the ballot but tells the software the voter does not wish to vote in that race.

The app would of course not store any images taken -- except the outside signature -- and securely erase them as well as any voting record.

Getting data

Volunteers would be asked to scan all variations of mail in ballots in the USA, particularly in areas of contested races where scrutiny is likely. Those uploaded scans would be quickly adapted to the system. Examples of validly marked ballots and poorly marked ballots would be added to the system to train it on this particular ballot and to recognize the sections of the ballot which need examination. Core training would be done for the common forms, such as checked box, or filled-in line, and for each region, tools would understand the local ballot and feed the boxes and lines to the general recognizer, as well as do OCR or other tools to recognize names. A tool would be built to make it easy for local volunteers to provide the information needed to analyze a ballot.

Volunteers could also make local videos for each different ballot about how to put the ballot in the envelope and where to sign etc.

In the future

This tool could, of course, be used in polling places that have a voter marked paper ballot, if it is legal to do so. Some places forbid photographing your ballot but this tool would look at the live camera image and never record an image of the ballot.

In addition, this tool, audited and open source, could make it much easier for voting officials to count paper ballots, but it's too late to do that for this election.

Trust

The software would be open source, and scrutinized for flaws and partisan bias by the community. In addition, it would be hosted by large trusted names and institutions to assure trust, and that the binary you get comes from the source that was inspected. Neural networks don't have "source" but they can be verified to a degree. Downloads would be signed and verified by the app stores.

Volunteers could also offer to loan their phones and help people check ballots, or donate old used phones to be ballot checkers. The risk here is concern if those ballot checkers are compromised. It would be nice if there were some way to make "official" certified checkers available at libraries, post offices in a way people without phones could trust.

While a cloud version could be done that only needs a digital camera or non-smartphone and a way to upload, it is harder to prove that this can be trusted, though it could be run by very high reputation non-partisan organizations. This would check your ballot