This is a great suggestion to keep in one's toolkit.

Obviously you cannot.

Not only to catch AI use, but also to better understand student deficits and provide their own substantive feedback, of course.

One of the fundamental problems here is severely overloading and underpaying adjunct faulty (or any faculty, for that matter).

This quote is from the comments. I think this is an excellent point about where LLMs (at least currently) fall down, and where a professor's insights and feedback is the "special sauce". And importantly, for professors, would you be comfortable not knowing whether or not an LLM had graded on something substantive vs surface level in a student's work?

I have had highly varied experiences with this, as some few professors I've had were clearly reading off slides or lecture materials they didn't prepare themselves, and didn't bother to review ahead of time.

It's absolutely true that supports like textbooks, slides, etc, are common, justifiable, and useful in a classroom. But again, you can purchase the textbook yourself, what has the professor added to make the course worth the cost of taking it?

Even worse! This is completely irresponsible teaching.

Even pre-LLMs, I remember taking an open book math test who's questions, when I examined them after the exam, were very clearly cribbed from practice tests published by a number of other universities (changing a few constants in an overall identical set of equations to be solved, for instance).

This approach made me lose a lot of faith in the quality of my instructors who couldn't be bothered to write their own novel math exam.

Yes, though I would also add "and have our teaching be responsive to the trends we see in student data, quantitative and qualitative".

And there's the natural consequence.

Again, the actual core problems here. Using AI to address this problem is putting the prof in danger of irrelevance. Anyone can write an (attempted) anthro paper and ask an LLM for grades, feedback, etc.

I think this is actually central to the problem. LLMs are being used to solve a problem rooted not in a need to access information, but rather to solve: - existing problems in higher ed, like attitudes towards who is responsible for student learning, and - newer problems in higher ed, like fundamental changes to funding which institutions are trying to solve by overloading class sizes (a solution that might not have been considered if there weren't already outdated opinions w/r/t the first point).

I think students' views that this is hypocritical may be reactionary, but it's up to professors to delineate, I think, how their ChatGPT use is fundamentally different from students'.

IMO, the use of LLMs to create novel works (course syllabus, answers to quiz questions, etc) robs the writer of a very crucial aspect of writing, the construction of new knowledge and analysis that's produced as part of the writing process.

Is this necessary to write and re-write a syllabus over each year? Probably not.

Is it necessary when writing quiz answers, essays, and other aspects of student work? Certainly. Importantly, students (and maybe some faculty) think that assessments need to reflect what is already learned and memorized, when in reality assessments provide more opportunities for learning though the process of assimilation and analysis of ideas.

Now, is this necessary for professors who are writing feedback and responses to student work? Yes. Using LLMs for feedbacks gets back to the central idea in this article; if the feedback comes from an LLM, what are students paying for? Importantly, how are professors factoring student work into their next lecture and overall into their instruction if they're not reading it?

Alongside this, is this necessary overall for professors? Probably yes. Lacking an understanding of what obstacles students are facing in their learning can only lead to an unresponsive class experience.

What is the special sauce students are getting out of a class?

If the professor isn't applying their personal experience and knowledge to their notes, what is the benefit of the student taking the class with that particular professor?

@chrisaldrich, @mrcolbyrussell, @steelwagstaff, @remikalir, @heatherstaines, @judell

We've quietly launched at-mentions in the Hypothesis web app. Yours were among the first names that popped into my head; feel free to suggest others (or at-mention them yourself).

Various marketing campaigns are coming, but I thought I'd ping you all here first.

test @mdiroberts

annotate in this open group!

You actually need /profile/settings.

LTI 1.1 grading result

annotation gamma

anno alpha

anno beta

test

page note

"TC" definition

LTI roles

LTI roles

Is this saying that Canvas' parameter User.id matches exactly the LTI parameter user_id?

Found under "Group Categories"

Found under "Group Categories"

Found under "Groups"

Found under "Courses"

Found under "Sections"

Found under "Courses"

Found under "Files"

Found under "Files"

Canvas scopes are, to me anyway, found in some unintuitive places. Creating annotations to help us find these again in the future.

Reason for the grades being 1-10.

Hypothesis is reporting the HTML of this selection contains a number of newlines that don't appear in the HTML on the page.

This is better known as the alt attribute!

$$\begin{pmatrix} 1 & 0\\ 0 & 1 \end{pmatrix}$$

$$\int_\gamma\|\gamma'\|$$

"Text feature" is defined in this document, but "text details" isn't. Giving students both definitions would help ensure they understand the author's intent.

Tags would also be a great way to practice metacognition! One technique would be to have a list of categories you use when annotating (such as "Evidence", "Central Idea", and "Vocab") and apply one of these as a tag to each annotation.

The goal is to go beyond analyzing the text and identify why you're analyzing that part of the text. You can read more about metacognition here.