Tools I Tried

Most of the tools on this page came to me from colleagues, from trainings, or from the teaching literature. I took each one far enough to see how it fit my courses: in some cases by building with it, in one case by watching it in a colleague's classroom.

MyOpenMath

Colleagues who already used MyOpenMath for CHEM V101 recommended it, and it is the tool I spent the most time on. I used it two ways.

CHEM V120B: pre-labs written and run. For the six-week Summer 2026 term I moved the pre-labs for all ten experiments out of the lab notebook and into auto-graded MyOpenMath assignments. Students had three tries per question, saw their score on each question immediately, and saw the answer and a short explanation after the last try. Grades reached Canvas through LTI, the standard link that lets Canvas open an outside tool and record its grades.

Questions in MyOpenMath are written as code, and at first I counted that in its favor. A question written as a script can be drafted by AI, so work I would otherwise do by hand could be handed over, and not knowing the platform's language was no obstacle. I used Claude to draft the code for each question from the lab manual and to check every numerical answer by calculation. The difficulty was getting the code into the platform. MyOpenMath has no API, meaning no way to send files to it directly, so the code for each question had to be copied and pasted in by hand, tested in the preview, and corrected. The format took five rounds of that before it held, and the other nine experiments were built from the first as a template. Browser tools that let an AI do the entry have since made the copying easier, though MyOpenMath is still comparatively unfriendly to work with.

CHEM V101: a full course, set up but never connected. In May 2026 I set up a V101 course for Fall 2026 from a colleague's existing course, with homework for Chapters 1–14 in 23 assignments. Before committing, I checked the homework against my ten quizzes, since my worksheets cover more than the quizzes test. The gaps were small, and I accepted them.

Decision. Students did not raise problems with it. The costs were on my side, in entering the questions, in editing a question once it was live, and in late work, where a student who needed more time had to spend a LatePass, the platform's built-in extension token. Sharing, one of the reasons I chose it, did not work the way I expected. A colleague warned me that an assessment copied from another instructor's course stays tied to that instructor's class, and a Canvas export of a MyOpenMath course carries only the links while the questions stay on the MyOpenMath server. Nothing I was asking students to do required MyOpenMath. After the summer term I did not continue with it, and I did not connect the V101 course.

The quiz feature in Canvas may be the better home for the pre-labs. It does the same job inside Canvas, any colleague can copy it, and it can be built from a script through Canvas's API, so I may move the pre-labs there.

Plickers

Plickers lets an instructor check understanding without student devices. Each student holds up a printed card with a code on it, and the way the card is turned marks their answer. The instructor scans the room with a phone or tablet and sees every answer at once. Seeing it run in a colleague's classroom, I thought it was a good tool, both in what it can do and in how smoothly it worked with a real class.

I looked at how I could use it in mine. It can be done. I already use my iPad to run and annotate my lecture slides, so scanning the room would mean a second device or handing the slides off for that moment, and I was unsure the extra step would be efficient enough in my lectures to try it right away. I have not used it in my own class so far.

In-lecture demonstrations

Joe Selzler suggested that I try demonstrations in lecture. When I asked him where to start, he lent me two volumes of Bassam Shakhashiri's Chemical Demonstrations and told me to look there. I also bought the American Chemical Society's Chemical Demonstrations: A Sourcebook for Teachers (Volume 2). The one I ended up running came from a video I found on YouTube while I kept looking: three flasks with the same amount of acid and increasing amounts of baking soda, and balloons that stop growing once the acid runs out. It ran in October 2025 and is described on the CHEM V101 Lecture Demonstration page.

I did not add more. Preparing a demonstration took real time, and stopping for one broke the flow of the lecture. It also made me settle how I divide lab and lecture. In my courses chemical demonstrations belong to lab, and in lecture I want the time to go to the lecture content. For lecture, the form of demonstration I settled on is the interactive simulations, which are on the CHEM V101 Interactive Simulations page.

PlayLab

In Spring 2026, through the college's AI Community of Practice, I used PlayLab to build a project guide for a CHEM V120B final project.

The project is long and has many steps. Students audit the claim that a protein structure is "known." They collect a corpus of 400-plus sources, draw one to three checkable claims from each, and verify the claims against experimental metadata and prediction-confidence metrics. In the first week, before any collection begins, they write inclusion and exclusion rules and a bias prediction, a statement of what they expect their own sources to over-represent.

The guide keeps track of where a student is in that sequence. It checks whether each step has been done, and when a student asks what to do next, it points to the next step. Its knowledge base is three documents I wrote: the course AI policy, the AI-use disclosure form, and the project handout. Following that policy, it helps with brainstorming, drafting and formatting, and it leaves interpretation and final conclusions to the student.

The guide and the project have not been used with students, and the guide remains a prototype. Four things kept me from taking it further. PlayLab does not integrate with Canvas. Using it would require an activity or assignment that makes students work inside PlayLab, and I was not convinced a course requirement should be built around a tool. It keeps no log of student use that I can collect, which makes anything students submit from it awkward to handle. Finally, the scale did not work. What I want from AI in a course is a project at a scale only AI makes possible, but as the project grows the model behind the guide struggles to handle it, and cutting the project down to what it handles well leaves little worth doing.