Randi Weingarten and the American Federation of Teachers MUST REVERSE Their Decision on ChatGPTZero

Or debate me | Or get sued | Or discriminate | Or have people think you are completely ridiculous

Oct 19, 2023

Stefan Bauschard is the author AI Writing Detectors Are Not Reliable and Often Generate Discriminatory False Positives,Co-Editor of Chat(GPT): Navigating the Impact of Generative AI Technologies on Educational Theory and Practice: Educators Discuss ChatGPT and other Artificial Intelligence Tools and the co-author of Beyond Algorithmic Solutions: The Significance of Academic Debate for Learning Assessment and Skill Cultivation in the AI World

Yesterday, I was sitting on the train working on a report on deep learning, and this article popped up on my LinkedIn feed -

I was dumbfounded. After reading the article, I was angry.

I reshared it with a couple of sentences of commentary.

It’s been re-shared 100+ times by leading voices in the growing field of AI education, including AI scientists, business professors, education professors, and global leaders in AI and education.

Ethan Mollick, one of the leading voices (60,000+ followers), on AI and education.

Yes, they are going to have to reverse this.

The Head of AI in Education at UNESCO also had a thought:

Comments like this were not uncommon. -

This one made me laugh —

These are the multiple reasons this announcement is a travesty.

(1) I’ve been cataloging ALL of the research on these detectors, and there is overwhelming evidence (this really isn’t even a debate) that these detectors are not accurate. That’s why OpenAI pulled their detector, why TurnItIn says to use it for nothing more than a conversation starter with students (which will be hard because, as Weingarten acknowledges, most teachers know very little about this technology), and why multiple universities have pulled the AI-writing detectors out of TunItIn. Anyhow, lots of false positives and false negatives ‘Just click on this link and look at the research yourself; don’t take my word for it.

(2) Students can easily slightly perturb the output to defeat the detectors.

I defeated one by changing two words.

They can have it add in spelling and grammar errors. They can write a few sentences and sentence phrases on their own. They can run it in and out of language translators. They can use a paid service to rewrite it to “sound” like a human.

(3a) The false positives often result from bias against students who are not native English speakers and have poor writing skills. Why? Because these detectors make percentage chance predictions based on the pattern in the analyzed text to determine if it was likely written by AI. Simple output patterns are more likely to be detected as “AI written” because AI output tends to follow simple patterns (this is explained in more detail here). Who writes in simpler patterns? Non-native writers and students who are poor writers. This obviously also falls most heavily on particular classes and races. You know, the ones who already bear the brunt of most of the discipline in schools. All that research is in my blog, but here are a couple of quick comments from reshares of my post.

What does AFT say about this? NOTHING. They don’t even acknowledge it. Are they aware? Do they care?

The AFT represents public school employees, and their mission is to: “champion fairness, democracy, economic opportunity, and high-quality public education, healthcare, and public services for our students, their families, and our communities.

They should actually try to do that.

Public schools are disproportionately filled with students who have lower-level writing skills and are most likely to be unfairly accused as a result of using these detectors, as they are with another discipline.

It’s nice to have photos on the website, but where’s the consideration for the well-known and unfair impact of the detectors on the people in the picture? Not a word!

3(b). Unsurprisingly, it’s also causing disciplinary referrals for special education students. Off Task: EdTech Threats to Student: Privacy and Equity in the Age of AI (September 2023)

(4) It’s the wrong approach. This is so sad because what teachers need to be doing is learning how to teach with this technology in the classroom. They need to learn how to reinvent assignments for an AI world. If they don’t, they are preparing kids for another century (and, yes, that’s been a problem for a while (this isn’t the teachers’ fault; they are told to teach a 100 year old curriculum).

Weingarten acknowledges that teachers need AI training, but where has AFT been on that? I can’t find a SINGLE thing about supporting teachers’ AI training on the AFT website as part of this initiative or otherwise. Their claim that they support students and teachers learning AI skills is hollow.

This is what they’ve been working on.

Before we go further, think of the narrative built by the titles of the first two articles. The first one seems to suggest parents and teachers want “guardrails” on educational technology, and the second one says AFT has a a control mechanism. I guess now we are supposed to follow what our parents say, which seems to be a shift in the AFT’s position related to other controversies.

But where’s the beef? There is none.

(1) There are only two sentences in the guardrails article about generative AI. The rest is about “ed tech” (as if this was just more “ed tech”) and social media.

(2) This is one of the sentences: “A majority of educators, 75 percent, reported that their students never use artificial intelligence software or AI-enabled search engines (e.g., ChatGPT, DALL-E, Bard) for learning. Likewise, 76 percent of educators say they never use AI software or AI-enabled search engines for their own work.”

How would they know? Plagiarism detectors don’t work, and just because students don’t use them at school doesn’t mean they don’t use them at home. How could one possibly go from asking teachers if students are using GAI to concluding they aren’t just because teachers say they aren’t? In their terms, it’s like asking the burglar if he stole. GAI is SnapChat and has thousands of applications! Do they seriously think kids don’t use Bard? This defies common sense.

Also, this is not consistent with other research. [Educators Battle Plagiarism As 89% Of Students Admit To Using OpenAI’s ChatGPT For Homework; More than half—58 percent—of students ages 12 to 18 have used ChatGPT, an AI-powered tool that can answer questions, write an essay on a Shakespearean play, or draft a legal memo that appears remarkably similar to what a human can produce, a Common Sense survey released May 10 found]

So where are they getting their information? Their percentages? It’s from a survey they paid for!

In the press release about this new AI writing detector program, they don’t mention that they paid for the survey (“A new Hart Research survey”) but they do provide a link. They also reference “The Surveys of Teachers and Parents on Educational Technology"—it's the same survey!

(3). This is another sentence: “The survey also shows that while both groups embrace education technology, 84 percent of teachers agree that teachers should have more say over how it is deployed in classrooms, and 55 percent of parents say they would like parents to have more say as well.” Note: There isn’t a single thing in this study that says parents want their kids to be punished based on plagiarism detectors. This is the AFT’s spin.

Hmm, it seems they commissioned a survey to support a narrative that teachers like “ed tech (see, they are actually progressive)” and want more control over it (is it that they are progressive or is it that they want to use writing detectors?), but that students don’t really use it very much (according to teachers). Parents want more control (they want writing detectors to be used?), so, hey, everyone wants this new program we made up, which is grounded in this survey that asks questions in a way that gets the answers we are looking for and is run in cooperation with recent college graduates.

Since AFT has 1.7 million members, I wonder why they didn’t partner with a larger and more established company such as TurnItIn, which is already in most schools and has an AI-writing detector (that many universities have turned off)?

I was pretty mad yesterday when I shared the article, but then I calmed down. Now I’m angry again. This survey nonsense is nonsense, and it gets even weirder (see below; they integrated a “misinformation” tool into the initiative).

What is going on?
__

(4) Lawsuits. Yes, penalizing students has consequences, and when it has a nonsense foundation, they sue you. This program is basically a lawsuit generator.

(5) We should spend the $ on training teachers about AI.

(6) It’s time to move on.

(7) Weingarten is terribly misguided.

After reading through some of the articles that have been written about this and the AFT’s announcement I’m honestly dumbfounded.

““AFT president Randi Weingarten told CBS MoneyWatch. "And the guardrails have to be about privacy and security and things like that." CBS News

How could running papers through plagiarism detectors possibly do anything but decrease privacy or security. And what are “things like that?” These last three words honestly suggest a strong lack of knowledge. Policies of this magnitude that attempt to impact every kid in every public school in America can’t be built around “things like that.” Did we send Ukraine some guns and “things like that.”

There’s more.

"We believe in its potential and we know if we don't guard against its perils upfront, we're going to repeat the terrible transitions that happened with the Industrial Revolution," she said. CBS News

I don't even have any idea what this could mean. Is a plagiarism detector going to stop unemployment and poverty?

(8) It’s bad to outsource your moral obligations

(9) This is intriguing.

Developed in January to scan text for AI input, GPTZero has since launched new tools, including one that allows students to certify their content as human, and to openly disclose when they use AI. CBS News

Certify their content is human? Why don’t the adults certify their own content is human? Hysterically (sorry, one can only laugh about this), AFT also promotes that teachers can use their integrated tools to generate non-human content!

It can hugely reduce paperwork burdens, bureaucratic burdens, and it can help with the writing of lesson plans," she said of AI technology. CBS News

The kids are done with a world where they are banned from using technology that is essential to their future while their teachers use it to do their own work. Has anyone noticed many kids have stopped coming to school because they are tired of BS like this?

Ironically, in a sad way, they can also go undetectable.ai and have it take their AI-generated text and make it undetectable, making it sound like a human. And then they can certify they didn’t “cheat.”

There is so much to certify these days. Does anyone have that form?

(9) In more weirdness, they’ve announced a partnership with NewsGuard as part of this initiative in order to stop misinformation. Huh? Yeah, misinformation is bad. I'm not sure what this has to do with plagiarism.

What is going on here?

Self-Immolation

In case you are wondering about my “self-immolation” reference, it is due to the fact that public schools are largely (not entirely; some (very few) are doing a few things) ignoring AI and the world we are entering.

Comparatively, private schools are doing a good amount of AI training. They haven’t blocked these AI tools (FYI: there are 1300 language models on Github that have been downloaded 600 million times since August, and some can run locally on your computer). I’m sure there is a survey somewhere about student use of these language models..

Every person who works at a job in the future (perhaps sans plumbers, according to Geoff Hinton) will use these tools at work. That’s the simple part of this technological transformation. More on all this on Monday, but we are at the beginning of the largest technological transformation in the history of the world, and it will all unfold over the next decade, with incredibly radical changes coming over the next few years. It’s time to prepare ourselves and our students for this. Sorry, but this transformation is not about plagiarism. No one invested hundreds of billions of dollars in tools to write school essays, and this is honestly the least of the challenges we face as a society.

A Positive Note

Pushing plagiarism detectors into schools is a really bad idea that needs to be reversed.

And whatever is going on with this survey should be shared. It smells rotten, to be honest. Combining this with NewsGuard (I have nothing against NewsGuard) makes it even stranger. Is the AFT evolving into some type of content-policing organization? I can’t figure this out for the life of me.

How is this positive? While AFT made a terrible mistake, they have so much amazing talent — 1.72 million members. If they were given proper professional learning opportunities as well as the chance to fully participate in discussions, challenges, and opportunities related to AI (no one has all of the answers, and I’m sure many of the members have brilliant ideas that could help), so much positive change could take place. Schools have so many programs—speech & debate, entrepreneurship, Future Business Leaders of America, Model UN, band, portfolio development programs—and classroom adaptations such as project-based learning, in-class performative assessment, interdisciplinary learning, robotics, etc, that could be utilized in such a positive way to help students succeed in this area.

And there are many good ideas within the area of writing.

Education needs to change.

Education and educators must prepare students for the new AI order of things. Educators lives are going to change in significant ways, not because their roles are likely to be automated away but because they need to teach a different curriculum and probably teach in a different way.
– Rosemary Luckin, Professor of Learner Centered Design at UCL Knowledge Lab

What is coming is a total change around how we really think about education, about everything that education involves, from the way a school looks like to the role of the teacher in the classroom.... ..Because the way that schools have been for the last couple of centuries has irreversibly changed and there is no coming back. Traditional educational institutions need to accept the disruption and fully embrace the change if they don’t want to disappear.
— Clara Lin Hawking, Head of AI @Globeducate

PS. I recognize a bit of this post is surely. I thought of deleting some of that, but I think it’s important that the frustrations come out. Like you, I’m human. We can all do better.

Debate

I’m always open to being wrong, and I like to debate. I’m happy to meet them anywhere to debate.

Some other thoughts

(1) Please keep this out of Australia