Watching Students Think: What a Librarian Saw When Students Started Working With AI
A teacher rarely gets to watch a student think. They see the finished product, but not everything that goes into it: the actual work, the wrestling with a topic, the moment something clicks.
Adoria Williams, Head Librarian and Department Chair at Merritt College in Oakland, found a way to see it. She built a Nectir AI Course Assistant into her research class, required her students to use it, and integrated their conversations into the assignment.
What she saw in those logs is something most educators never get to see: the shape of a student learning to think. And once she could see it, she could teach into it.
Watching students’ interactions with AI change
When Adoria first introduced her Nectir Course Assistant, “Bixby,” she polled the class to see where they stood on AI. Most were somewhere between wary and nervous. A few admitted they'd used AI to cut corners in other classes. Several were afraid that using it would get them in trouble. The most common early reaction wasn't excitement. Adoria's students didn't know how to talk to an Assistant at all. She kept getting the same confused response: "What do I say to it?" Her answer:
"Talk to it just like you're talking to me. This is your tutor. It's built right into the class. Have a conversation."
To get them started, Adoria used a mix of Nectir-provided prompts and her own prompts designed specifically for her course. Then, as the weeks went on, she pushed them to change the prompts themselves.
"I would provide the prompts, and I would encourage them, as we move through the course, to add to the prompts, so that they can learn to formulate questions that will get the results that they want."
A yes-or-no question gets a yes-or-no answer, and a student walks away no smarter than when they started. A real question, one with context and a point of view attached, forces a real exchange. That progression is visible line by line. A student who opened the term typing a flat request was, by midterm, giving Bixby real context, stating a position, and asking it to challenge that position back. Bixby would ask what they thought, whether they'd considered another angle, whether they had an opinion of their own.
"And students were responding."
You can't see that shift in a finished essay. You can only see it in the record of the conversation that produced it.
The skill you can actually watch students build
The habit Adoria cares most about is the one that happens after the AI answers. A student reads what comes back, and then does something with it.
"Students have to remember to evaluate, verify, revise, and make their own decisions when using it."
In a research course, the point isn't collecting answers. It's learning to interrogate them, to recognize bias, and to notice what's missing from a source before you rely on it. Adoria puts it plainly: a key skill in information studies is learning "to recognize biases and to look for the missing context." Most students don't arrive knowing how.
"They don't understand that they have to put the final touch on whatever AI gives them by looking at it critically, and double-checking, and looking for things that could be erroneous."
Nectir is built to support those skills rather than paper over it. An Assistant can show the sources behind its responses, so a student can follow a claim back to the material it came from instead of trusting it blindly. For a class whose whole subject is source evaluation, that turns every exchange into a small exercise in the thing being taught.
When the conversation becomes the work
Adoria still grades everything herself, by hand, with detailed feedback. But one thing she grades has changed. Students’ conversations with Bixby are now part of what students turn in. While Nectir does not give instructors access into exact chat logs to protect student privacy, many instructors like Adoria integrate it into their teaching, having students export and upload their conversations with assignments.
Take one of her assignments. A student searches the library database for a print book on their research topic, writes a citation for it, and then brings both to Bixby for feedback.
"They need to use Nectir AI to give them feedback about that citation. Does it look right? Is anything missing? And then I might have them use it to give feedback on the choice of the book they selected, whether it will support their research question."
Because the Assistant is grounded in her course materials, that feedback stays tied to what she actually taught, not to whatever a general tool would improvise. The student's judgment is on display the whole way through: which book they chose, how they framed the question, what they did with the feedback they got. The transcript captures the reasoning, not only the result. For a course built on teaching a process, being able to grade the process is the whole point.
"Showing and training students how to use it intentionally, I think it's key."
Why watching learning matters
Administrators worried about AI rarely worry that it can produce text. They know it can. What they don't know is whether students learn anything while using it.
Adoria has the receipts. She can point to a first-week transcript full of "is this good?" and a final-week transcript where the same student is defending a narrowed research question against pushback, and she can show the upward slope between them. That's learning, documented, on the exact tool people worried would replace learning.
None of it happened by accident. It happened because Adoria set the Assistant to coach rather than answer, required enough interaction to make it real, and then paid attention to what came back. The transcripts weren't a surveillance tool. They were the clearest window she'd ever had into her students' thinking, and what she saw through it was them getting better.
Want to see what AI support looks like when it's built into the tools your campus already uses and controlled by your faculty? Schedule a demo, and our team will walk you through how it works at schools like yours.
Frequently asked questions about Nectir AI
What is Nectir AI? Nectir AI gives colleges, universities, and high schools the ability to deploy AI Assistants across their campus that are fully controlled by faculty and administrators, built into existing learning management systems, and compliant with FERPA and SOC 2 Type II standards. Nectir is trusted by 150,000+ students across 130+ campuses, including a partnership with California Community Colleges, which serves 2.2 million students across 116+ campuses.
Can faculty see how students use Nectir AI? Yes and no. Individual conversations stay private, and a student's exact chat logs with the Assistant aren't visible to their instructor. What faculty can see through analytics dashboard is the bigger picture: the general themes and topics a class is asking about, or areas of confusion. Faculty can access chat logs by integrating conversation exports from students into assignment submissions.