From Mean Drafts to Keen Emails

LLMs make it OK to be blunt and rude!

[image created with Dall-E 3 via ChatGPT Plus]

Welcome to AutomatED: the newsletter on how to teach better with tech.

Each week, I share what I have learned — and am learning — about AI and tech in the university classroom. What works, what doesn't, and why.

In this week’s piece, I report on some stats about student and faculty use of AI, I discuss how LLMs can take you from a blunt and rude draft to a perfect email, and I briefly reflect on Google’s snafu.

📢 Quick Hits:
News Tidbits for Higher Educators

  • Temporarily, Google suspended Gemini’s ability to produce images of people after it was found to produce inaccurate depictions of people from history, like German soldiers from WWII or American founding fathers, in ways that represented diversity where it didn’t exist.

    • Why it matters: We all know about how AI tools can hallucinate or otherwise produce content that misrepresents reality, but many critics have interpreted this case of hallucination to reveal the ideological bias of Google. It may. However, it is the sort of side effect we should expect from AI companies that incorporate more substantive ethical views into the mechanics of their tools. Google certainly differs from other more laissez-faire AI companies in its approach to AI — in general, they are slower to release tools, more driven by their views on how AI ought to be used, and quicker to make adjustments in light of criticism. It remains to be seen whether companies that operate like this are taking a better approach to AI in the long run.

  • The University of Pennsylvania’s School of Engineering and Applied Science will now be offering “one of the very first AI undergraduate programs in the world, the B.S.E. in Artificial Intelligence.”

    • Why it matters: Gen Z students are especially sensitive to the utility of their degrees and are looking for ways to stay abreast of the AI revolution, so it is going to be crucial for universities to experiment with offerings like this.

  • Adobe released the beta version of its AI assistant in Acrobat, which enables you to interact with pdfs to ask questions about them, summarize them, and navigate their content.

    • Why it matters: The pdf format has been a hard nut to crack for tools like ChatGPT because many pdfs are designed to be read by humans, not machines. It is intriguing to see the biggest player in the pdf market engaging with this issue through its popular software, especially if Adobe allows integration with other suites and ecosystems.

💡 Idea of the Week:
Many Students are in the Dark

There is more and more evidence that students are using generative AI like ChatGPT for their school work. While we all have anecdotal evidence of this phenomenon, survey results corroborate.

A recent study in the United Kingdom of 1200 undergraduate students by HEPI and Kortext found that 53% of students have used AI to help them prepare their submissions for assessments, with 36% using it as a “private tutor.”

Another study by ACT Research of 4000 high school students in the United States found that 46% had used AI tools. It also found that AI tool use correlates with ACT score (more tool use correlates with higher scores, though AI tools cannot be used during the test) and that the majority of students who use AI use it more than once a month.

On the other hand, many faculty are slow or cautious to engage with the developments in AI and the developments in their students’ preferences. A study this past fall by Tyton Parners (commissioned by TurnItIn) found that faculty use AI at half the rate of college students — namely, they found that ~50% of college students use it but ~25% of faculty use it.

However, if you are a faculty member reading this, then it is likely you are ahead of the curve. Anecdotally, we have found that faculty who are AutomatED subscribers are much more open to using AI — or are at least more actively engaging with its challenges and opportunities — than faculty more generally.

This brings me to my idea of the week: be cautious about assuming that your students are aware of even the basics of the AI space.

Although the average student is quicker to adopt AI than the average faculty member, many students have no clue about it at all, so those of us who are trying to stay ahead of the curve can mistakenly operate under the assumption that our students are advanced in their AI knowledge and AI skills.

For example, as I worked to train my students in building and using custom GPTs in my Philosophy of AI class this past week, I found that some already had built custom GPTs to help them with their schoolwork or to have fun, while others were truly shocked at what custom GPTs can do. And these are students in my class!

My experiences match with the results of the above studies. The second study cited above found that 83% of those students who haven’t used AI tools are “not interested in using them,” so it is no surprise that many students have little awareness of their nature. The third study cited above found that, “apart from 12% of students identifying as daily users,” most students’ use cases were “relatively unsophisticated” like summarizing or paraphrasing text.

For those of us in the AI-curious bubble, we need to continually work to stay current, but we also need to recognize that what we take to be “common knowledge” is far from common outside of the bubble.

🧰 An AI Use-Case for Your Toolbox:
From Mean Drafts to Keen Emails

In our Premium piece yesterday, I explained how to better integrate Gemini with Gmail to respond to emails you receive in Gmail that are answerable by information in your Docs.

By default, there are a few options to use Gemini to help enhance emailing — using Gemini independently or within Workspace via the “Help me write” tool — but each leaves something to be desired.

I showed Premium subscribers how they can take it to the next level by leveraging the greater power of Gemini in Docs. Unlike Gemini in Gmail, which is limited to generic requests and cannot reference old emails or files, Gemini in Docs can produce responses sensitive to the specific information within a given Doc. My solution automatically moves the content of emails that users designate in Gmail into the appropriate Doc and adds space for them to use Gemini to draft responses to them. After drafting responses based on the content of the Doc and the moved email, they send the drafts back to Gmail.

Although this more advanced solution is quite powerful, it bears emphasizing that the default “Help me write” functionality in Gmail will still be useful for some use cases. For instance, if you want to spend less time responding to basic emails where your response would be very short if you could be rude and blunt, then Gemini in Gmail will be more than capable.

However, the emphasis is on “will be.” It seems that Gemini has been significantly restricted in its power over the last few days since the controversial snafu noted above. Hopefully, it will be back to its prior form — issues aside — by the end of the coming week.

Still, we can illustrate the power of this mean-to-keen strategy with Microsoft 365 Copilot. (Before I get any farther: if you have no clue what Microsoft 365 Copilot is or how to get it, click here to learn more.)

Take the case of a student email like this one:

Suppose that your answer, in a nutshell, is this:

yes you have to pay and it’s worth it

Type this blunt and rude response into the “Draft with Copilot” window…

…and here is the result:

That is, Copilot in Outlook produces this:

Dear Jane,

Thank you for your email and for bringing up your concerns about the ChatGPT plus subscription. Yes, the subscription is $20 a month and it is a requirement for our class. I understand that this may be an added expense, however, I believe that the tool is well worth the investment. The ChatGPT plus provides advanced capabilities that will greatly enhance your learning experience in the course. I have found that students who use it have had a much better understanding of the material and have performed better in their assignments. I hope that you will consider obtaining the subscription and make the most of this opportunity. If you have any further concerns or questions, please feel free to reach out to me.

Best regards,

Graham Clay

Not bad!

Beyond the integrated LLMs like Copilot for Microsoft 365 and Gemini for Workspace, or the dedicated apps like Superhuman, you could duplicate this functionality with some copy-pasting and Claude or ChatGPT.

I guess the only question is whether using this strategy will reduce you to a jerk in everyday conversations. I am prepared to risk it…

Have you tried out our course design GPT? Give it a try, if you have ChatGPT Plus! It can produce assignments, assignment sequences, rubrics, and AI course policies. We have designed it to be especially effective when it comes to pedagogical issues related to AI.

Remember, you can even get it involved in any other GPT conversation you are having, if you @ it!

We just got our first rating this week after OpenAI rolled out the rating system We are constantly working to improve it, so please give it a rating to tell us how it performs.

👀 What to Keep an Eye on:
Our April Webinar (on AI Training?)

Two weeks ago, we hosted our first AutomatED webinar. This one was 2 hours long, covered custom GPTs in depth, and enabled our participants to begin to build their own. Some of our attendees were interested in GPTs to help with their own pedagogy, some were creating GPTs to assist other educators, and some were interested in ways to ease access to their companies’ content.

We are already planning the next webinar. We are thinking it will be one hour long and occur on a Saturday at noon Eastern Time in April.

If you are interested, please answer the poll below so we have a better sense of which topic(s) we should cover:

What topic should we cover in our April 2024 webinar?

Login or Subscribe to participate in polls.

🔜 What’s Next for ✨Premium Subscribers

Late in the fall of 2023, we started posting Premium pieces every two weeks, consisting of comprehensive pedagogy Guides, in-depth Tutorials of AI use cases, releases of exclusive AI tools like AutomatED-built GPTs, and Q&As with the AutomatED team.

Our next Premium piece will be released during the week of March 11th, and it will cover how to use AI tools to improve your workflow around virtual meetings with Zoom, Google Meet, and Microsoft Teams.

So far, we have six Premium pieces:

To get access to Premium, you can upgrade for $5/month or $50/year, or get one free month for every two (non-Premium) subscribers that you refer to AutomatED. The price will be going up in March.

To get credit for referring subscribers to AutomatED, you need to click on the button below or copy/paste the included link in an email to them.

(They need to subscribe after clicking your link, or otherwise their subscription won’t count for you. If you cannot see the referral section immediately below, you need to subscribe first and/or log in.)