All or nothing at Educause24

At On EdTech+ on Friday we covered additional context on NSC enrollment data and misunderstandings of the demographic cliff. On Saturday we covered two studies on generative AI’s impact on learning and the funding of noncredit programs. On Monday we covered Coursera’s changing content strategy that builds on content produced by the company itself. Today we welcome back guest author Kevin Kelly.

Full disclosure: As you’ll see, Kevin (with me providing minor support) facilitated a partners workshop for Educause in a paid engagement.

As Phil Hill noted in a recent Friday Follow Up, he and I went to San Antonio to facilitate the first Educause Annual Partner Summit, attend the Educause conference, and enjoy some Texas BBQ. During the conference I spent several hours each day in the Educause Exhibit Hall, specifically to identify notable trends among the vendors solution providers.*

It's all (still) about artificial intelligence (AI)

Looking beyond AI roadmaps

It's been almost two years since OpenAI launched ChatGPT and the rapid rise in the use of generative AI tools in the mainstream. Over numerous AI-related posts (and with numerous grains of salt), Morgan and Phil have commented how companies are or are not incorporating AI into their product roadmaps, how different stakeholders are reacting to AI-related hype, and how higher education in general faces challenges when it comes to AI.

For my part, over the past couple of years I've been working with higher ed students, faculty, staff, and campus leaders to consider a balanced approach to AI. Sometimes this takes the form of modules for my online class, keynote talks, or faculty development workshops (e.g., see "What AI prompts would Darwin write?" or "Integrating GenAI into your pedagogy"). In late 2022 through 2023, the concerns were simpler. On one hand, educators want students to submit original work without using AI. On the other hand, educators want to prepare students for using AI as part of their future work experience. Now teachers and academic technology professionals have more robust questions, like "Do EdTech platform roadmaps consider campus policies related to AI?", "How well do AI chatbots help students find support services?", and "Can I build a custom AI bot to support my students?"

With these types of questions in mind, at Educause I spoke with representatives from companies that are creating AI pathways for teachers and students to perform academic tasks, such as using AI to make sense of course topics (BoodleBox) or support the assessment process (TimelyGrader, Kritik). Here are some notable trends:

• Alignment with campus policies: Colleges and universities are worried about how AI tools will align with their current or upcoming campus AI policies, especially those related to privacy, data use, etc. (NOTE: University Business reported in September that only 3% of campuses had AI policies.) Every solution provider confirmed that the work done by students and teachers would not be used to train the AI engine/LLM (or at least would have the option to opt out).

• Choose your own AI adventure: Several solution providers (e.g., BoodleBox, QuadC) have created tools that not only work with a campus' LLM of choice, but also allow users to toggle between different AI engines such as Claude and ChatGPT. This allows users to compare results and use the "best tool for the job."

• Integrate AI throughout a workflow: Whether it be student success (QuadC) or assessment (Kritik), the providers have thought through how to integrate their AI tools throughout a workflow. In some cases, users query the AI; in other cases, the AI generates information for the user to review and act upon. I was impressed that Kritik included a calibrated peer review activity in its assessment workflow, which I've done with my own students for years. Something like this would definitely simplify the process.

• Moving from prompt engineering to bot building: Some tools like QuadC allow instructors to create their own bots to support students with completing specific learning tasks. Stanford's dSchool created Riffbot to allow educators to integrate reflection into assignments. QuadC allows non-programmers to create bots for to address specific academic support goals.

• More complex problem-solving: AI chatbots (e.g., Gideon Taylor's Intrasee or ivy.ai) have gotten vastly more sophisticated. Trained on a college's entire web presence, now they can help end users with complex tasks. For example, if a student needs to fill out a form to resolve an issue, the chatbot will help them fill it out and submit that form without leaving the chat. In some cases the user may not even realize they've completed a form, having only answered some questions prompted by the chatbot.

It will be interesting to see how each company handles the costs associated with making AI-related queries. Some companies are pursuing subscription models, including a subset that plans to offer exemptions for Pell-eligible students.

Nothing about ghost students and fraudulent enrollment

Explaining fraudulent enrollment

Not all AI news is good news. In particular, AI has exacerbated the problem of fraudulent enrollment--i.e., rogue actors who use fake or stolen identities with the intent of stealing financial aid funding with no intention of completing coursework. First these 'ghost students' apply to colleges and then register for classes if their applications get accepted. If they can stay in a class until the census date (the last day students can be dropped for non-participation), then they get a financial aid check. A few years ago these criminals might at least get caught by the class census date. Now the scammers are submitting AI-generated essays to avoid being dropped for non-participation.

The consequences are very real, including financial aid funding going to criminal enterprises, enrollment estimates getting dramatically skewed, and legitimate students being blocked from registering for classes that appear "full" due to large numbers of fraudulent enrollments. If you haven't heard of this problem, then you also haven't heard how big the problem has become. Last year the San Francisco Chronicle reported a notable example: 20% of California Community College applications -- 460,000 out of 2.3 million -- were fraudulent between July 2022 and June 2023.

The challenge for higher ed involves balancing the need to both a) verify the students are who they say they are and have legitimate intent to complete courses and programs and b) create an equitable ID verification process. When describing what would make the process more equitable, educators provide a variety of reasons ranging from supporting students without documentation to avoiding small, incremental burdens like students paying for a notary public or having to go to campus. Increasingly, an additional burden will fall on legitimate students to prove their identities if they want to take online classes. State-wide college and university systems have been investigating multi-layered approaches that include different tools to filter spam, verify identification, stop bot traffic, and provide an IP (Internet Protocol) quality score.

Searching for solutions at Educause

In this year's Exhibit Hall, Educause set up a "neighborhood" for Cybersecurity, Privacy, and Risk Management, although I didn't see Mr. Rogers there. I spoke with between 15 and 20 companies focused on cybersecurity and/or identity management, and only half had even heard of fraudulent enrollment. Almost all of them said that their work focuses on securing the environment after users' identities already have been verified. A handful offered concrete leads, but none of the companies had a solution for any aspect of the identity verification process.

• Law enforcement angle: One rep suggested finding a solution that taps into the Criminal Justice Information System. It's not clear if any solutions listed below do this.

• Rumor mill: One rep believed that Docusign had a tool that would allow students to upload documentation to verify their identity. I was able to confirm that Docusign does have an ID verification feature.

• Emerging tools: Tools being used by other industries, such as 1Kosmos and Verosint, may offer solutions for higher ed.

• Existing tools: One rep referred to IDme, a tool that some community college colleagues have investigated.

Based on the some preliminary research, the Educause team might consider recruiting some additional Exhibit Hall participants: A.M. Simpkins (S.A.F.E.), BMTX (Identity Verification) Human Security (Account Protection package), IDme, Mitek Systems (Verified Identity Platform), and/or Verosint (Account fraud detection and prevention).

The truth lies somewhere in between

Perhaps it's fitting that I saw Everything, Everywhere, All at Once on the plane ride to San Antonio. There is a multiverse full of issues to solve and, although I logged a lot of time in the Exhibit Hall, it's only one Exhibit Hall of many. Regardless of which booth I visited, I approached each person and framed my questions as someone trying to solve a specific problem. Every solution provider rep gave me his/her/their full attention, asked for more details, pulled in colleagues, and provided helpful suggestions. Many gave honest answers when their product or service would or could not do the job. In other cases, they would turn to their product and demonstrate how it was a possible solution; however that was not the first thing they did.

This brings me to the asterisk (*) from my opening paragraph. At the Annual Partner Summit, we discussed how the word vendor -- stemming from the Latin word vendere ("to sell") -- immediately casts industry partners in a sales-focused light. In her closing remarks, Leah Lang from Educause proposed using terms like solution provider instead. The hours I logged in the Exhibit Hall showed that the EdTech industry appears ready to engage with educators in this way. We just have to approach education-industry conversations in a different way.

The main On EdTech newsletter is free to share in part or in whole. All we ask is attribution.

Thanks for being a subscriber.