danaleeling

As part of an introduction to MS 150 Statistics I showed the students how to enter a few basic statistical functions into a spreadsheet . I then used a Gemini notebook (formerly NotebookLM) preloaded with the textbook chapters to check the work. I started with a prompt: I want to give you some data followed by some statistical questions along with my answers. I want you to check my answers. The data: Car speed in mph 21.2 21.2 . . . 21.2 20.0 The questions: Statistic to calculate Sample size n Mode Median Sample mean x̅ Answers 18 21.2 20 20.3 Note that in the above the data was clipped, the original prompt included all of the data. Gemini/NotebookLM responded with the following analysis: Gemini notebook correctly discerned that the data set was multi-modal. There is a mode.mult function that can be used to return values in a polymodal data set, but that is not usually covered in the course.  The takeaway is that Gemini notebook correctly parsed and responded to a multipart prompt...

"What is 120 plus 12?" I asked the student. The student and I were measuring the circumference of a hoop that exceeded the length of the 120 inch tape measure. The student had chosen to measure the circumference and to use inches. The tape measure was 12 inches short of the circumference, the student was holding the tape measure showing that the gap remaining was 12 inches. The student looked up from the measurement in hand and said nothing. I waited.  Students in a prior term timing the period of a hoop. The period of a hoop is linearly related to the diameter. "Just add the 12 to the 120, what is 12 more than 120?" The student looked at me blankly. I waited. After twenty seconds, the student asked, "I can use my cell phone calculator?" The student pulled out their cell phone and obtained 132.  "You couldn't do that calculation in your head?" I asked. The student indicated that they could not.  ❖ "Plummeting math scores prompted lawmake...

This summer physical science will again pilot AI feedback in Moodle. This summer the laboratory reports are dropping the explicit rubric and using essentially the same marking system but without an explicit rubric. The original rubrics for the course appeared as above circa 2010.  December 2016 By 2018 the rubric used in Schoology had shifted to a ratings scale to conform to the way rubrics are set up in a learning management system. With a shift in ratings terminology and points, this structure would carry over into Instructure Canvas. March 2021 statistics rubric built on outcomes in Canvas Canvas allowed course learning outcomes to be included directly in the marking rubric. Physical science rubrics also included content criteria. January 2026 in Moodle By 2026 rubrics in Moodle had become minimalist. Experience had indicated that students were not reading nor reacting to the rubric marking. Where in Instructure Canvas learning outcomes could be included in rubrics, in Moodle le...

I do not use AI detection systems in SC 130 Physical Science for laboratory report submissions as I have reasons to doubt their accuracy. AI is evolving too quickly for detection systems to keep up. In addition, false positives are deeply damaging to the instructor-student relationship. False positives are undoubtedly very rare, but anecdotal evidence suggests that false positives disproportionately affect the most academically capable students.   The course syllabus does not currently include a statement on the use of AI because as an instructor I am still working to understand what a functional AI policy will look like. I can only encourage the students to write reports in their own words. My intent is to help them improve their own writing skills, not the writing skills of an AI.  Sometimes I can sense that AI has been used - AI tends to use vocabulary and idioms found in formal scientific writing and technical publications. That is what the AI is trained on, but my st...

"Princeton will now require instructors in exam rooms for the first time since 1893" - The House of El . A former Yap campus instructor, used to say, "Tests don't test." The college president has encouraged the development of curricula that is inclusive of AI. I still have tests and quizzes in my courses, but they are done unproctored, outside of class, on the students' own time, in Moodle. The tests are obviously open book and open world. The tests are also not used to assess learning. More complex assignments and projects are used to assess learning.  I spent time during this between-terms break consuming videos on AI in education and upcoming developments in AI. I can see around me the issues El describes in the video: instructors, and institutions, trying to find ways to retain an educational structure that predates 1893. Proctors, lock down browsers, paper and pencil in class exercises and assessments.  A study on how to redevelop the college curricula ...

Upon uploading a video , YouTube now transcribes the audio, runs that transcript through Gemini, and then produces prompt suggestions as questions in a dialog box at the upper right. The dialog box appears if one clicks on the Ask dialog below the video. The video title and description are now trimmed to half the video width to accommodate the Ask dialog. This is not a feature I enabled. This happened automatically after uploading my video. ​ Some sources online indicate that a related or similar feature may have been made available to YouTube Premium users a year ago. At Google I/O roughly two weeks ago Sundar Pichai said "It's been 10 years since we pivoted the company to be AI first. We knew how profound it would be to advancing our mission."  I have seen a video clip with outtakes from Google I/O that indicate that a related feature, Ask YouTube, is new. As I do not recall seeing the Gemini Ask prompt in previous uploads, I take this to be a server side update that...

"Future jobs depend on being better than AI" Born after the advent of LLM AI he will grow up in a world infused with AI and will never have known a time before AI.  What if humans cannot be better than AI? What future jobs can humans do in a world where humans cannot be better than AI? Watching a presentation on education in this age of AI the presenter recommended aligning education with future employment where humans are better than AI. The presenter also recommended having students analyze AI where responses had gaps and cracks. The history of science and religion teaches that a God of the gaps is problematic. AI is in its infancy, the hallucinations, gaps, cracks, and errors will shrink if not disappear. An OpenAI internal model just solved an open problem in discrete geometry through reasoning, not brute force. AI will correctly solve any mathematics problem. Why would humans need to spend decades learning to do something that AI has learned to do in a frac...

In Nature Volume 652 from 16 April 2026 on page 543 Chris Bivins argues that the world needs more taxonomists. One particular line in the article caught my attention, "But when I find a fungus that lacks a name, I can still hold it, examine it under a microscope, compare it to known species and begin the work of description. An LLM cannot. It has no access to the physical world, only to text."  That LLM's only have access to text is simply not true. LLM's are now quite proficient at image analysis. Give the LLM images of known fungal species and the LLM can begin the work of description, potentially highlighting similarities and differences to known species. I gave Google Gemini app this image: And the prompt "Describe this photo along with the likely identification of the organisms in the photo." The response started at the blue diamond: This image captures a close-up look at a brush-footed butterfly resting amidst lush green foliage. Visual Description The...

April 15  Laboratory reports in two courses are marked with rubrics. This is an exploration of the capabilities of AI assisted feedback ( AIF ) in Moodle 4.5. The AIF plug-in was configured in practice mode in assignments for both courses.  The first example is from SC 250 Botany, a thin layer chromatography analysis of plant pigments. The laboratory report is marked using a rubric as seen below.    The prompt given is seen below along with setting up Generate feedback automatically on submission for Practice mode, which is the formative feedback mode for AIF.   For AIF to use the rubric, the Grading Method must be set to rubric. For Practice mode, the marking workflow must be set to No, which effectively turns off the practice mode. There is a summative Teacher-Reviewed mode that involves a different set of settings.         The following is an example from SC 130 Physical Science Ohm's law laboratory exercise. This laboratory is also...