What Is "Human Intelligence," in Proposed Initiative Ballot Title, as Opposed to "Computer or Artificial Intelligence"?

An interesting opinion by the Arkansas AG's office, rejecting a proposed popular name and ballot title for a draft initiative.

In Arkansas, as in many other states, proposals for initiatives' popular names and ballot titles must be submitted to the state AG's office, which must evaluate them to make sure they fairly and accurately summarize the initiative's legal effect. The AG's office may then accept the popular name and ballot title, revise it, or reject it.

An Arkansas AG's opinion released last week (No. 2023-108) rejects the following proposal:

Popular Name

An amendment to the Arkansas Constitution to conduct all elections by paper ballots containing inherent security features which protect the integrity and authenticity of an official ballot, with vote selections marked by hand using permanent ink placed directly on the ballot by the voter (except when otherwise required by federal law), permitting the counting of election day votes only after the polls close on election day, requiring the vote count to be verified by human intelligence before certification of the vote, requiring that all elections in this State be conducted by voters selecting only one candidate or issue per race with the winner determined by which candidate or issue receives the majority plus at least one vote of the total votes, preserving the special runoff system, ensuring that elections cannot be conducted in this state using an internet, Bluetooth, or wireless connection, and allocating funding to ensure free, fair, and secure elections.

Ballot Title

An amendment to the Arkansas Constitution that repeals Amendment 50, § 4 and amends Amendment 50, § 2 to remove the language that permits elections to be conducted by voting machines, modify the language of Amendment 50, § 2 as well as adding new sections to Amendment 50 which effectuate a policy and practice that all elections in this State must be conducted with paper ballots containing inherent security features designed to prevent unauthorized duplication, with vote selections marked by hand using permanent ink placed directly on the ballot by the voter in a manner which ensures the secrecy of the votes cast on the ballot and the anonymity of the voter casting the ballot, requiring that no election day votes be counted or tabulated before the close of the polls on election day, requiring that the tabulation of votes be verified by human intelligence before certification of the vote, requiring that all elections for government positions or issues in this State be conducted by voters selecting only one candidate or issue per race with the winner determined by which candidate or issue receives the majority plus at least one vote of the total votes, ensuring that elections cannot be conducted in this state using an internet, Bluetooth, or wireless connection, and allocating funding to ensure free, fair, and secure elections….

The AG gives several reasons for rejecting the proposed name and title (laid out in detail here), but here are some that struck me as especially interesting:

• "Human intelligence." The popular name, the ballot title, and the text of the proposed measure all contain the phrase "human intelligence," which the text defines as "the thought and physical process of a human being instead of the thought or process of a computer or an artificial intelligence." This language is confusing because the meaning of "thought and physical process" are unclear. Does this mean a human cannot use any machine or computer, such as a calculator or Excel spreadsheet, to aid in his or her "thought and physical process"? The answer to this question would surely give voters "serious ground for reflection." [Under Arkansas law, {sponsors cannot omit material from the ballot title that qualifies as an "essential fact which would give the voter serious ground for reflection."}]

The answer is also important to determine which statutes would be supplanted by the amendment. For example, under A.C.A. § 7-5-602(c), paper ballots must be "run through an electronic vote tabulation device before a hand count is conducted." And A.C.A. § 7-5-606 specifies the requirements for "exhibit marking devices and electronic vote tabulating devices." Since your definition of the term "human intelligence" is unclear, I cannot ensure that the ballot title is not misleading….

• Secrecy of individual votes. Amendment 50, § 2 to the Arkansas Constitution currently requires that the secrecy of individual votes be maintained: "All elections by the people shall be by ballot or by voting machines which insure the secrecy of individual votes." The text of your proposed amendment would repeal the secrecy requirement in Amendment 50, § 2. Section 6 of your proposed text would replace Amendment 50, § 2 with the following language, none of which maintains the current "secrecy of individual votes": "All elections by the people in this State shall be by a paper ballot containing inherent security features which makes the paper ballot difficult to duplicate or counterfeit." This repeal might also be read as having the effect of repealing the provision of Amendment 81 to our constitution that protects the secrecy of votes.

Yet your ballot title summarizes the proposed text as "ensur[ing] the secrecy of the votes cast on the ballot and the anonymity of the voter casting the ballot." You appear to believe that the method of voting required by your text would itself ensure the secrecy of individual votes. That may be true. But since it is unclear to me whether you intend to remove the constitutional right to the secrecy of individual votes, I cannot ensure your ballot title is not misleading….

• Popular name length. The popular name—at 152 words—is longer than a typical popular name. It instead reads like a second ballot title. Although this alone is not misleading, you may wish to significantly shorten the popular name to better meet the purpose of popular names as described above.
• Partisan coloring language in the popular name. It is my opinion that your proposed popular name contains impermissible "partisan coloring" language when it uses the word "integrity." The Arkansas Supreme Court has held that "partisan coloring" language is "a form of salesmanship" that "gives the voter only the impression that the proponents of the proposed amendment wish to convey of the activity represented by the words." The word ""integrity," as used in the popular name (i.e., "protect the integrity"), gives voters only the impression that the proponents of the proposed amendment wish to convey—a "[s]teadfast adherence to a strict moral or ethical code … being unimpaired; sound[]." To paraphrase the Arkansas Supreme Court, the "[voter] is entitled to form" his or her "own conclusions" on whether the proposed measure promotes integrity.
• Partisan coloring in the ballot title. It is my opinion that the ballot title also contains impermissible "partisan coloring" language when it uses the words "to ensure free, fair, and secure elections." Such words, like a slogan, give voters only the impression that the proponents of the proposed amendment wish to convey— as if to vote otherwise is to ensure the opposite of those characteristics. Again, the "voter is entitled to form" his or her "own conclusions" on whether the proposed measure promotes elections that would be ""free, fair, and secure." …
• Grammatical issues. Your ballot title only uses commas. But because of the length and complexity of your ballot title, which includes multiple instances of a series within a series, the use of semicolons would provide greater clarity and promote readability. Additionally, random capitalization appears throughout the text of your proposed amendment, which does not appear to serve any purpose. You may wish to correct this.

Comments

[Rev. Arthur L. Kirkland]

Unless I am mistaken, the link indicates a destination of Arkansas Attorney General Opinion 2023-108 but connects with Arkansas Attorney General Opinion 2023-109.

[Rev. Arthur L. Kirkland]

Is the Volokh Conspiracy going to (1) correct its error without acknowledging the mistake or the correction or (2) let the error stand as a long-term tribute to its level of scholarship?

The record makes it a coin flip.

Carry on, self-described “scholars”

[Brett Bellmore]

It really is important, when you're trying to do a ballot initiative, to have it checked over thoroughly for mistakes like this.

As a side note, I really wish Congress had to live with those "partisan coloring" rules for bill titles...

[Rev. Arthur L. Kirkland]

This was submitted by someone whose website indicates he was named "one of the top three lawyers in Saline County."

I have a few hunches about Saline County but this guy's work speaks for itself.

[Krayt]

Top 3 = #3, or else he'd say top 2. Top 2 = #2 or else he'd say number 1.

[DaveM]

The “intelligence” part needs to be stricken in its entirety. There is no such thing as actually artificial “intelligence”, that’s just a popular name we use to refer to the output of a type of search algorithm. Any intelligence in that processing is entirely human in origin. The machine follows instructions and makes selections based on weightings and tunings of parameters.

Likewise the unintentional striking of security requirements. Oof dah.

[SRG2]

I think that some neural networks are genuinely artificial intelligences when they can add layers and connections.

[DaveM]

Yes, they "add layers and connections", but 1) only at the highly specific prompting of an actual intelligence, and 2) only build on information that was already created by people with actual intelligence.

What AI will return to us are "surprising" results, but not intelligent ones. There are human-curated tunings built into the algorithm to discard the most outrageous/useless/false search results, but there is no underlying intellect in the algorithm. You can see this most clearly in AI-assembled images of animals, where frequently there will be too many AND too few toes and fingers. The algorithm has no concept of "animal", it just has image data associated with the word, and in some of those images toes are hidden and in others, shown. Since it has no knowledge, it just picks among the most likely next-image-blocks randomly, sometimes erring too many, sometimes too few, sometimes just right.

But it's still surprisingly useful because we, as actually intelligent beings, are easily able to discount the nonsense and noise and recognize the image as a "cat". What will continue to surprise us is the speed at which it can generate recognizable output for us. That will be truly revolutionary.

Sorry for the long post!

[Zarniwoop]

This proposal sounds like it was written by someone who has never volunteered as a poll worker, most notably the part “requiring that no election day votes be counted or tabulated before the close of the polls on election day”.

The municipality where I live uses “fill in the bubble” style paper ballots and an electronic tabulator. The voter fills in the desired bubbles, walks over to the tabulator, and inserts. The ballot is scanned and (if there are no errors like undervote, overvote, illegible markings, etc.) the device goes “bing!” and informs the voter that their ballot has been tabulated; if there is an error, it spits the ballot back out and the voter can correct the mistake.

If accepted, the ballot itself drops into a bin at the bottom of the machine and is kept. The machine reports the counts after the polls close; since the ballots are kept, they can be manually recounted if needed (which is a great way to check on the tabulator software!). The system is easy to understand, and each voter can directly, visually satisfy themself that their vote has been counted and that their ballot is safely stored for any recount.

This initiative language seems to preclude this type of very secure, paper-trail-verifiable vote tabulation, which is sheer madness for several reasons.

1) If ballots can’t be fed into a tabulator while the voter is watching, voters will not like that; there will be complaints that poll workers could alter or replace ballots after the polls close.

2) Related to #1, tabulation only after the polls close presents logistical problems for vote security and integrity. Who’s got custody of election-day ballots before tabulation? How are they stored? They can’t just be piled up on a table in the basement of the local church.

3) Running ballots through a tabulator allows a voter to correct any mistakes or illegibility on their ballot. Tabulation only after all the voters have gone home eliminates that, which will end up necessitating a lot more “well, what do we think the voter intended?” judgment calls by election workers.

4) Permitting tabulation only after the polls close means workers – volunteers in many jurisdictions around the country – will spend significant extra hours feeding ballots into a tabulator (and Dog help us all if it’s interpreted to require manual, human-eyeballs tabulation with a pencil and paper).

[Zarniwoop]

The requirement also effectively eliminates another security check that occurs continuously throughout election day: the ability to make sure that the vote count (voters who have signed in and been handed a ballot) matches the number of votes tabulated. Even 1 extra vote would get noticed and receive a lot of attention. And since there are lots of eyes on the tabulator at all times, no one can just walk up to it and slip in a few extra ballots without being noticed.

In contrast, if someone slips extra ballots into the stack awaiting tabulation after the polls close, and there are 1100 votes but only 1000 voters who received a ballot ... how do you notice that until the stack is tabulated? How do you fix that? This initiative would make stuffing a ballot box easier, not harder.

[Longtobefree]

I agree, the fill-in-the-bubble with automatic verification when inserted is the best system. It allows for straight-forward human verification, and easy recount by machine or humans. After the polls close, it can then transmit totals for each machine for aggregation and final results reporting.
Of course, this is the system democrats hate the most.

[apedad]

Man, it must be hell in your head.

[Brett Bellmore]

The scantron system is superior on so many metrics it's hard to understand why it isn't almost universally adopted.

And, yes, this badly drafted proposal would prohibit using some of it's best features.

You could still have ballot checking, without tabulation on the spot.

Requiring that the ballots be directly marked by the voter, and be human readable, this is perfectly reasonable. Prohibiting machine tabulation? Not so reasonable.

[Rev. Arthur L. Kirkland]

.

This proposal sounds like it was written by someone who has never volunteered as a poll worker,

Yeah, but he's one of the top three lawyers in Saline County, Arkansas . . . just ask his website.

[Savagely Average]

Any idea what prohibiting vote counting until the polls close is supposed to accomplish, even in theory?

[santamonica811]

Savagely,
Sure, that's easy. It ensures that the vote count will be delayed. Thereby permitted any butt-hurt losing candidate to claim fraud and/or irregularities.

Now, if you were really asking, "What supposed *benefit* will this give us?" . . . well, then I'm as bemused as you are.

[Savagely Average]

OK, that was pretty much my understanding as well. Yeah, it seems really stupid even just from a Republican partisan perspective (I'm assuming a Republican is sponsoring it). As you say it can potentially help a candidate who loses a very close race to contest the result. But there's no reason to assume such candidates are more likely to be Republicans.

[Don Nico]

SM,
I interpret that clause to preclude, opening and tabulating votes before election day. SImple counting of submitted ballots would be allowed.

[Absaroka]

Just a wild guess, but didn't some of the historical examples (cough...Chicago...cough) involve the ballot stuffers watching the count through the day, then delivering just enough votes to get their guy over the top?

I guess this implies an assumption that stuffing ballots is easier before the polls close than after. Whether that's true would depend, I suppose, on the details of how the polls and counting are done. If, for example, you know 10463 ballots have been cast before you start counting, it would be hard to add the 200 fraudulent votes towards the end of the counting process.

Just guessing, though.

[David Nieporent]

.

Any idea what prohibiting vote counting until the polls close is supposed to accomplish, even in theory?

I don't know, but I can speculate: conspiracy theorists argue that election fraudsters wait until they see how the vote is going to decide whether and how many fraudulent ballots to submit. So if you count the votes early, they know they need to submit exactly 2,851 fake ballots; if you wait until after the polls are closed to start counting, they can't do that.

[Brett Bellmore]

I agree that the proposal is badly drafted, but the intent of this provision IS fairly clear, and Absaroka and David have identified it: To deny would-be fraudsters advance knowledge of how many votes they have to manufacture.

[Zarniwoop]

Brett sez:

So if you count the votes early, they know they need to submit exactly 2,851 fake ballots; if you wait until after the polls are closed to start counting, they can’t do that.

I guess that's a possible intent, but it doesn't make sense in the real world for a well-designed system. Good process design isn't a partisan thing ... but beware of bad process design based on partisan fear-mongering and conspiracy theories.

In our (arch-liberal Madison WI) system, the tabulator counting the scantron ballots doesn't report the election results until after workers close the polls and start printing the official tally. There's a real, old-fashioned paper tape! And a phone modem to upload/report the same results to a central location. The paper tape and sealed bags of the actual paper ballots are also hand-carried to the central location where they can be manually verified and recounted if necessary. Poll workers can also print out as many spare copies of the paper tape as they want for souvenirs/shennanigan detection.

That's a key point: the scantron tabulator is a stand-alone machine, and the results can't be read either locally or over a (non-existent) internet connection until after the polls close. It reports only the number of ballots cast (for cross-checking with the voter rolls, a solid check on voting ghost people).

So conducting the tabulation process as voters mark and cast their ballots on election day, which confirms to voters in real time that their scantron ballot has been accepted and is in the (locked) ballot bin, doesn't provide information necessary to "know[] how many votes they have to manufacture." It. Just. Doesn't.

Also: absentee votes are also tabulated during the day; there is no opportunity to add absentee votes after the polls close.

Bad cases might make bad law, but conspiracy theorists make worse law.

[Brett Bellmore]

"I guess that’s a possible intent,"

It, expressly, IS the intent, if you've ever spoken to the sort of people with these concerns.

Who are not just Republicans; The people behind this proposal actually are concerned about Republicans stealing elections!

[Absaroka]

Are you sure? Your link seemed non-partisan, but this doesn't sound like the initiative proponents are middle of the roaders.

[DaveM]

The reason for not counting until polls close is to eliminate first-mover advantage. Say you count votes every day. You have a leaker on your elections staff who sees a trend running against you in Precinct #1, but well ahead in #2 a week ahead of election day. They advise you to stop spending any time and money to get out the vote in #2, and transfer it to #1.

You just threw the election.

[Archibald Tuttle]

I choose to vote Libertarian. Last year, my 3 page ballot showed a long list of party names I never even heard of, but only one Libertarian candidate for 1 office. I left all the others blank. So I practiced massive undervoting. Should my ballot have been rejected?

[Zarniwoop]

No, of course not. That’s a more detailed case I didn’t feel a need to get into, but since you asked, the process at the scantron tabulator looks like this:

1) voter inserts scantron ballot
2) tabulator beeps and says “hey, you undervoted, do you want to try again, or accept as-is?”
3) tabulator shows “cast my votes” and “return ballot to me” selections
4) voter decides what to do

If the voter accepts the undercount, the ballot is tabulated, and dropped into the sealed and locked ballot bin. All votes and non-votes are reported at the end of the day just like any other ballot.

If ballot is returned, you can correct undervotes as much as you want. Or not. There’s also a comparable process for correcting overvoted ballots.

Not rocket surgery.

[Archibald Tuttle]

Thanks. The devil in in the details.

[Zarniwoop]

No problem. Yeah, there will of course be some scenarios like that.

Markings that the scantron tabulator considers extraneous overvotes are far more common.

So the ballot gets spit back out, the voter tears their ballot in half in front of a pair of poll workers, and requests/receives a new one. The interaction is logged. The replacement ballot is voted/tabulated, and the spoiled ballot is also kept (there's usually a handful per election in the precincts I've worked in - not large, but non-zero).

It's the level of detail that's ... already a solved problem.

[DaveM]

I have been serving as a volunteer poll worker for a couple of decades now, and have used several different types of voting machines. I completely agree that Scantron is far and away the best method for counting, recounting, and auditing ballots. It is robust, accurate, verifiable, easy to administer, and easy to use.

But all methods of counting have weaknesses, and a paper-based system is no exception. Scantron voting is most easily attacked by stuffing paper ballots you want counted, and intercepting paper ballots you don’t.

What makes this a “good” weakness is that it’s physical. You have to physically gain access to the ballots you want to add or suppress. That puts it miles ahead of a digital ballot, which can be attacked at a distance, and, depending on the method used to convey the data, potentially attacked en masse.

Because paper requires physical controls to establish a chain of custody, I am definitely in favor of in-person voting, whether early or late. We should only allow the paper to follow a completely secured path from the voter to the tabulator. I am not in favor of mail-in voting, that is a horrible reduction in security and reliability. I am also not in favor of very long terms of early voting, every day that you have to secure paper only adds to the difficulty. My own state unfortunately allows both mail-in ballots AND has a very unsafe term of early voting (45 days).

But these flaws are easily corrected. Allow mail-in voting only for people who will be physically absent. Get rid of drop boxes, require voters to appear in person for voting, whether early or on election day. Absolutely guarantee that ballots go directly from a voter’s hands into a tabulator.

If I ran my state’s elections procedures, that’s how I’d do it, anyway.

[Zarniwoop]

DaveM says:

But all methods of counting have weaknesses, and a paper-based system is no exception. Scantron voting is most easily attacked by stuffing paper ballots you want counted, and intercepting paper ballots you don’t.

This is exactly why I'm against the proposed law as drafted: it would appear to prohibit the voter feeding their own scantron ballot into the tabulator on election day, immediately after marking it at the polling location, so that there is zero chain of custordy beyond "in the voter's hand". Instead, it would require that a voter hand their ballot to election workers, with a "trust us, we'll feed the ballots into the machines after y'all go home" mode of operation. Voters would, I think, correctly call BS on that as a step backwards, not an improvement.

As noted above, "tabulation" is not the same as making a running tally of the election results available in real-time, prior to polls closing. A properly-implemented system that tabulates at the precinct level during the day with results reporting only after polls close eliminates the purported issue of "knowing how many votes are needed".

[Stephen Lathrop]

Note also that the language seems to mandate first-past-the-post, and to outlaw constitutionally any sort of ranked choice or other alternative. Seems like that is worth mention in the summary.

[Don Nico]

It does not appear to forbid ranked choice but only to discourage it.

[Brett Bellmore]

"requiring that all elections for government positions or issues in this State be conducted by voters selecting only one candidate or issue per race with the winner determined by which candidate or issue receives the majority plus at least one vote of the total votes,"

Sure looks to me like it outright prohibits ranked choice. You can't do ranked choice if voters can only make one choice per race.

[DaveM]

Ranked choice voting is the only method of counting that ensures that no one gets their first choice, and in a very close race, that the winner will make no sense to anyone. /sarc

[Kleppe]

They should definitely choose a different term of art.

https://en.wikipedia.org/wiki/Human_intelligence_(intelligence_gathering)

[Dilan Esper]

Isn't "human intelligence" the antonym of "signals intelligence"?

[Don Nico]

Yes, it is a separate intelligence control system, HUMINT

[Michael P]

Not an antonym, but HUMINT is a complement to SIGINT. Other similar categories include MASINT, IMINT and GEOINT. RUMINT doesn’t really count….

[Kleppe]

HUMINT is just RUMINT that you have to pay for.

[ReaderY]

The Attorney General’s critique does not go far enough. Pens, papers, chisels, clay or stone tablets, all are kinds of machines. Only by using the human voice alone can human thought be communicated without the use of machines. Any permanent record of any kind is mechanical in nature.

[santamonica811]

Well, I'd add American Sign Language to the example of using one's voice. 🙂