votingworks / arlo

GNU Affero General Public License v3.0
141 stars 25 forks source link

Add RCV data entry interface for audit boards #1308

Open mcchilders opened 2 years ago

mcchilders commented 2 years ago

Description: For audits of the RAIRE RCV Ballot Comparison type, the audit boards should see a slightly different ballot data entry interface that allows them to indicate ranked votes for various candidates. See screenshot here: Screen Shot 2021-10-19 at 11.46.48 AM.png

Changes to this interface from our normal one include:

cjerdonek commented 2 years ago

I was reading the suggested interface, and it looks like it might not be able to represent things like—

I wonder if the ranking dropdowns should instead be multi-select (though you would need to introduce more vertical space to display multi-values). Alternatively, the UI could be transposed so the user would select a candidate (or "overvote") for each numbered choice. The former may be better though, especially for write-ins, because it would better match how the real ballot is laid out (at least for hand-marked paper ballots).

(I guess it might depend on the jurisdiction's rules. If the jurisdiction would count the example above as 2nd choice Paul rather than 2nd choice overvote, then the UI wouldn't need any changes.)

mcchilders commented 2 years ago

@cjerdonek yep, it depends on the jurisdiction's counting rules - but good catch, because we do need to allow multi-select in the dropdowns since, for your example, Rank 1 would count but Rank 2 would be an overvote -> exhausted ballot under SFs rules.

@MorganLove note the addition ^^ in the description; a shift/select should allow you to select multiple ranks within a single dropdown (which is an edge case/error, but one we need to account for.)

cjerdonek commented 2 years ago

@mcchilders Another idea occurred to me that I think would make data entry a lot easier and may also simplify the implementation, including removing the need for multi-select and displaying and storing multi-select, etc. You could add an "Overvote" choice below "Write-in." Then, in my example, it would be Sandy 1st and Overvote 2nd. Data enterers would only need to indicate the highest ranking for each candidate and, if there is an overvote, the highest ranking at which an overvote occurs. One reason this would simplify data entry is that e.g. if someone ranks Sandy 1st, 2nd, 3rd, 4th, 5th (e.g. at all rankings), which isn't uncommon, data enterers wouldn't feel like they need to select all 5 since the UI wouldn't permit it.

(By the way, I'm fairly certain assigning a candidate multiple rankings isn't normally called an overvote. It's only when more than one candidate is assigned to the same rank. So you may want to update your overvote bullet in the top description.)

mcchilders commented 2 years ago

@cjerdonek you're right that it might simplify a lot of things, but one of the core principles in RLAs is that you want auditors to enter what they see on the ballot, the votes the voters actually marked (subject to whatever voter intent guidelines are used by the jurisdiction, of course); you explicitly don't want auditors interpreting those voter marks according to how they think the tabulator would have/should have counted them. So we never ask anyone to decide, for example, that something is an overvote, and we wouldn't want them to only mark some votes and not others (e.g. only the first ranking.)

Re: terminology for multiple rankings for the same candidate, I'm sure you're probably right, but I haven't found another term for it anywhere (and I've looked!) Let me know if you come across one.

Given that the instructions given to SF RCV voters are:

I feel ok calling both of those overvote conditions (although the first exhausts the ballot and the other just skips you to the next ranking under SFs rules).

cjerdonek commented 2 years ago

@mcchilders Okay, thanks. That's a valid position regarding the manual entry. Out of curiosity, how would the audit be affected if the ballot says Sandy 1st and 2nd, but the data enterer only enters Sandy 1st? Could it result in more ballots needing to be sampled because it's a strict mismatch, or would it never cause more ballots to be sampled because it would never affect any of the vote totals?

umbernhard commented 1 year ago

Finally picking this ticket back up again. @cjerdonek, to answer your question like a year and a half later, it depends on the audit type. In ballot comparison it could blow up the audit if Sandy was the runner-up who suddenly gained a valid vote. Or it could do nothing, if Sandy is the winner. In a ballot polling audit it's less likely to have a large impact, again all depending on rulesets and whether Sandy is the winner or not.