Dr. Bruce Forciea

AI Integration /Author/Educator/SME

Update for 2024-2025

 

This chapter includes an update of my experiences using AI in my classes over the 2024-2025 academic year. Once again, please excuse the imperfect writing since this is entirely human written. І thoroughly enjoy writing and am now beginning to feel sorry for those who offload their writing to AI since І believe reflective writing is one of the greatest intellectual activities, plus, it’s fun!

І am presently using MS Word and to the very left of this sentence is the icon for Copilot which will willingly write this chapter for me if І so chose. It is somewhat tempting to offload my work to AI, but then what am І actually gaining by producing content in the recognizable AI style? Where is the reflection, organization of thoughts, and process of describing concepts in a unique style when letting AI do the writing for you? І firmly believe the writer always learns something from writing and the process of developing a written piece stimulates thoughts and improves thinking. І see the words “actually gaining” in the second sentence of this paragraph were highlighted by Word and when І click on the blue dotted line under these words, Copilot pops up wanting to make my writing “better.”

І often think of students sitting down to do homework with such tools providing a constant temptation to do their work for them, offering the choice to complete an assignment in minutes versus hours. І also see the endless and unrelenting competition for attention from students’ mobile devices which І believe further fuels the temptation to offload work to AI. The combination of the easy use of AI with the learning deficits resulting from the Covid pandemic and the continuous dopamine secretions from social media interaction results in a “perfect storm” of opportunity to just let AI do the work.

І read articles about education and AI every day and recently I’ve been seeing more mention of how students perceive general education courses as unnecessary since AI can do just about anything required in these courses. The thinking is along the lines of “why should І do this if AI can do it?” This subsequent decrease in motivation to complete coursework coupled with the fact that even if you spend the hours necessary to complete assignments, you may end up with a lower grade than classmates who used AI also contributes to the dilemma of AI use educators must deal with.

Like us, students have busy lives and where І teach there are students with families and jobs pursuing second careers. І give them a lot of credit for wanting to improve their education, and І can relate since І worked full time while earning my first degree. І can see where it would be very difficult to ignore these wonderful magic tools that complete assignments in minutes.

 

Cognitive Offloading

The struggle against what is known as “cognitive offloading” was one of the challenges during this last school year. Cognitive offloading refers to the act of letting AI do your thinking for you. І witnessed a very large increase in students using AI in my classes for just about everything. The use increased dramatically from Fall of 2024 to Spring of 2025.

Specifically, what І have seen is an increase in students reflexively reaching for their devices to answer questions during class. In the past, more students attempted to recall information covered in previous class sessions or in the lecture completed just a few minutes ago. Some even used the notes they had just taken from the lecture. Now, it seems as though the default method is to immediately go to AI.

For years, І have recognized the importance of communicating with students about the relationship between what they learn in class to their future careers. They want to know why they need to learn something and what the value of learning will bring to their future careers. During this past year, І increased my development of activities and assessments that directly relate course content to future careers in each class meeting and constantly remind students why they need to learn the content while completing these items. 

І have worked to discourage the reflexive use of AI during class and have developed more interactive lectures whereby students would help me with the lecture via a Socratic type of questioning. For example, if we were studying heart anatomy, І would tell students to put away their devices and ask simple questions like “what is the job of the heart?” or “where is the heart located in the body?” І would use these simple questions as a starting point to move into more detailed heart anatomy. The idea was to move more into a process of learning versus just answering questions. Students could then build on what they already knew.

Of course, interactive lectures using this approach are not easy, especially compared to presenting PowerPoints, writing prewritten notes on the board, or a technique I’ve seen others do and don’t agree with which is reading from the text in class. They do, however, provide a great opportunity for teaching versus merely providing information. In the past, educators were the source of information but now, especially with AI tools, teaching can enter an entirely different dimension consisting of application of knowledge and analysis of ideas.

I encouraged AI use in my classes and have integrated AI into my activities and assessments, so a lot of what І had seen was expected. І feel that it is extremely important to create a safe environment for students to use AI if they wish, which is great because І can then observe how they use AI while wandering around the classroom while they are working on group discussions. My class also begins with an open discussion about recent developments in AI and AI use. 

І have observed some generational differences regarding AI use as well. One observation from the Fall of 2024 was related to an activity whereby students apply concepts they worked on in homework assignments during the past week to case studies in class. І noticed that younger students tended to immediately go to AI for answers while older students accessed their homework. This tendency to access AI for answering questions increased dramatically between Fall of 2024 to the Spring of 2025. One example was a student who wore earbuds during class and І suspected he recorded the lectures in real time and was able to immediately answer any question І posed to the class. Others very quickly accessed AI on laptops and mobile devices to answer simple questions.

This led to my increased use of Socratic teaching methods whereby students were required to present their findings verbally during class while І asked questions about their findings after their presentations. І attempted to require them to think on the spot during these presentations. The semesters began with some students struggling to do this but after a few weeks they adapted and improved dramatically to the point of attempting to anticipate my questions. 

Much like the Copilot icon continually tempts me to offload writing this chapter to AI, students’ continuous access to AI tempts them to offload their thinking to AI.

 

Improving Custom AI Chatbots

Development of custom AI chatbots and GPTs was covered in a previous chapter, but І wanted to cover how these were improved over the Summer of 2024 and during the semester break before the Spring of 2025 semester.

І firmly believe that AI can be used to improve learning, especially some of the more difficult concepts in my Anatomy and Physiology courses. My use of custom chatbots began in the Fall of 2023 with a simple chatbot used exclusively for struggling students needing a tutor, since none were available. The tutor ran on GPT 3.0 and was successful in explaining concepts at various levels. І continued developing custom chatbot tutors for use in the Spring of 2024 semester with a large increase in student use during this time.

During the summer of 2024, І revised my main custom chatbots to provide shorter answers to user prompts and to increase interactivity. For example, using the generic ChatGPT 4.0 results in a lot of content for each prompt which І find students are reluctant to read. The other issue is that even if you ask for references, some of these may be hallucinated, making it difficult to track where the information comes from.

The summer 2024 revision included altering the master chatbot prompt to provide shorter amounts of content along with choices for the user regarding what they want to learn about next. For example, our topic of heart anatomy would be broken down into several choices consisting of chambers, valves, vessels, etc. The user then makes a choice and the chatbot provides another smaller section of content and asks the user if they want more detail or move to another topic. The goal was to provide smaller bits of content and increase interactivity.

The chatbot was successful in both goals of smaller content and increased interactivity and students reported they liked the improvement. The chatbot was still trained on course content and provided references to where the information came from.

One use-case for the course chatbot was to do role playing activities which mimic patient interactions with health care providers. For example, the chatbot could role play a patient with diabetes or congestive heart failure. The student could interview this patient and provide strategies for managing the disease.

І also developed a Socratic Tutor chatbot that begins by asking the user what topic they want to learn about and then asks very simple questions that progress to more complex questions in a very interactive way. This chatbot was developed in OpenAI using GPT 4.0 which incorporates advance voice mode whereby students can speak to the bot on their computers or mobile devices.

The third chatbot developed in the summer of 2024 was a more advanced hospital drama game aimed at nursing students. The chatbot does a great job of creating an entire drama complete with characters, dialogue and dramatic cases where students must make choices regarding answering questions and providing treatments to patients. І used this during class as a game activity and to help stimulate critical thinking.

Deleting Exams

One discovery during the break between the Fall 2024 and Spring 2025 semester was about how AI could now easily answer questions with anatomical images. І had tested this capability back in the summer of 2024 and found that it could not correctly identify structures on complex anatomical images as well as some special questions І developed with diagrams of processes. This changed during the Fall of 2024 whereby AI was now able to correctly identify not only images of anatomical structures but could also do this in camera mode using a phone within seconds.

І was using these exams since Covid since everything had been converted to online courses but had been decreasing the points and making them more formative while increasing points on my AI integrated reflection assignments. І ended up deleting my exams after this discovery.

This meant that new unit assessments needed to be developed, and this had to be done over the short break. І decided to develop some very complex cases with the help of AI that focused on unit learning objectives. The cases were to be completed in 2 parts. The first part was during an in-class meeting whereby students worked through the class in small groups and presented their findings subject to my many questions and the second part consisted of writing a human written reflection describing what was learned from the case.

It was required for students to participate in the class discussions for these complex cases. І worked with student groups throughout the meetings by helping them focus their learning on specific aspects of the course and relating the content to what was already covered in the unit.

This proved to be successful during the Spring 2025 semester, although І did observe nearly every student using AI to solve the case problems. The cases were purposely made to be very difficult and on the level of a medical college course, but it was really great to see students understanding some very complex concepts by the end of the activity. І could assess their knowledge via their oral presentations and their answers to my questions done without AI use. My hopes were that this approach could continue to be used in future semesters. The main point was to use AI to reach higher levels of understanding.

This is something І could never have done in the past when the focus of these courses was on learning basic information. AI allowed students to apply and analyze complex concepts to reach greater levels of understanding.

More Efficient Chatbots

One issue that surfaced during the Fall 2024 semester had to do with students running out of credits for free AI accounts in POE or OpenAI. My chatbots were running on GPT 4.0 and using a good deal of credits for each message which resulted in quickly depleting their credits.

Then, right after the first of the year in 2025 the AI world was again disrupted with the release of DeepSeek which was a much more efficient LLM. My thought was that if DeepSeek can create very good responses to prompts, І may be able to use a more efficient LLM to drive the chatbots. My chatbots also obtain information from uploaded documents, which decreases the “thinking” required by the chatbot since information is pulled from the docs.

І experimented with a Llama model from Meta and a GPT mini LLM from OpenAI to drive some course chatbots which dramatically decreased the number of credits by around tenfold. Both worked good but Llama produced a few rare hallucinations and did not directly reference course content while the GPT mini model did. The hallucinations resulted in some great class discussions about the accuracy of AI generated content and potential bias.

Then came the release of GPT 4.1 mini. This LLM was better than the previous mini LLMs and used much less credits. І used this LLM to power another version of the course chatbot and polled my students regarding whether they ran out of credits with free accounts. І was happy to learn they could use this chatbot extensively without running out of credits. Problem solved!

 

Student Progression of AI Use

During the Fall of 2024, І observed most students were using AI. There were a few holdouts, however, which is okay with me since І have designed my courses so that students can choose to AI or use the course content (text, videos, interactive activities, lectures). Students use AI during class group discussions with one or more students in a group accessing AI on their laptops or mobile devices.

During the Spring 2025 semester, I noticed that every student used AI at some time or another. For example, in my advanced anatomy classes, everyone used AI during complex case assessments.

Students improved their use of AI as a learning tool throughout the school year with some students demonstrating sophisticated prompting and excellent interactions with AI. Overall, prompting improved and non-disclosed use of AI decreased dramatically, even in the online course.

Typically, at the beginning of each semester, some students will attempt to “test” the use of AI generated assignments. Some are very obvious with recognizable AI generated content. І use a “placeholder zero” grades for these if undisclosed and improper AI use is suspected. Students can either redo the assignment or meet with me to go over their process for completing the assignment.

І am aware of the inaccuracies of AI detectors, however І recommend students use some of the free detectors to check their work. І explain to them that we are living in a Post-AI World and we are working to preserve learning and reducing cognitive offloading. There are some students who use more sophisticated techniques such as embedding special characters using white colored font into assignments, using misspelled words or AI humanizer software to avoid detection.

І’m sure that not all incidences of undisclosed and unethical use of AI are detected and it is becoming nearly impossible to recognize AI generated content. Sometimes students will just use a different prompt or pivot to a different LLM to make their work read as if written by a human. 

The result can be a kind of paranoia in instructors in which one doesn’t know if the work was completed by the student or by AI. Sometimes this leads to a forensic approach where instructors use multiple detectors, scan text versions of assignments and closely analyze writing to catch perpetrators.

І must admit that sometimes І have fallen into this mode, especially since І work very hard at teaching the use of AI as a learning tool and not a homework tool. Occasionally I need to take a deep breath and step back from forensic investigations. Fortunately, my in-class sessions helped to dissuade my fears about student learning. We will never be able to catch all perpetrators, and we must remember education is ultimately an individual’s responsibility.

What helps is to regularly discuss the proper use of AI along with topics including cognitive offloading and using AI as a tool for learning versus for doing assignments. Toward the end of the Spring 2025 semester, several students stated that they were unaware of the capabilities of AI as a learning tool before taking this class. One student said that he had used AI in high school for cheating, but this was the first class in which the appropriate use of AI was addressed as a learning tool. І took that comment as a win!

During the final unit assessment in my Advanced course І noticed that every single student had a laptop and was using AI to help them solve the very difficult multisystem case. This was the largest use of AI І had observed over the past 2+ years. It was great to see what they discovered and І was sure to encourage them to explain the concepts, relate concepts to previous coursework and develop their own hypotheses about the case. My hopes are these activities increase learning through the application of course concepts to cases they may see in their future careers in healthcare.

 

Use of Camera Mode During Class Activities

Students are increasing their use of Camera Mode during class activities. For example, ChatGPT’s camera mode allows you to take pictures with your cellphone and will then identify, analyze, or summarize the content in the image. Students’ use of this technique increased over the past school year. By the end of the Spring 2025 semester a good deal of students would take pictures of the whiteboard after lectures and then let AI produce notes, explanations and reports from the images.

One of the activities in my General Anatomy class is for students to use portable whiteboards to draw diagrams of the circulatory system. The diagrams are then checked by me for accuracy. Students are allowed to make corrections until the diagrams are correct.

І decided to attempt to use AI to analyze the diagrams for accuracy and with the students’ permission took a picture of their diagram with my phone in ChatGPT’s camera mode. І prompted it to check the diagram for accuracy and produce a score on a scale of 1 to 10.

ChatGPT did an excellent job of analyzing and scoring the diagram. Other students saw this and wanted me to get AI to analyze their diagrams as well. І did and they were enthusiastic about the results. The class then became a competition about which group could score the most points from the AI analysis.

І plan to encourage students to do this during the next semester. This is a great use-case for using AI as a learning tool by getting feedback before the diagram is checked by the instructor.

In some cases, my interactive lectures consist of concept maps about a topic students have covered before the class meeting. Students first produce concept maps in small groups on the portable whiteboards and then present these to the class. My lecture then consists of producing a master concept map with the help of the class.

After the master concept map is completed, І take a picture and prompt AI to produce notes from the image. І can then post the notes in the course’s online LMS. Now, more students are taking pictures for their own notes.

Plans for Summer 2025

І am working on a plan for the upcoming Fall 2025 semester regarding how І will be using AI in my courses. І plan to continue using a course chatbot and will redesign it to further increase interactivity so that students can complete a learning session with the chatbot and copy and paste the entire interactions into assignments. The GPT mini model will continue to be used unless an even more efficient model emerges from the AI scene.

The new chatbot will be even more closely linked to course learning objectives and will ask students which module they are working on. The chatbot will then teach the content in the module.

The prompts І had previously provided to students have evolved over the past 2 years. They began with a learning objective style and then developed into topics.

Examples of Student Prompts:

2023: Learning Objective Style Prompt:  “Describe the anatomy of the brain”

This was a more general prompt directly related to a learning objective. Students would enter the prompt and the chatbot would provide the information at any level of understanding.

2024: Topics Related to Learning Objectives Style

  

Topic: Brain Anatomy

Subtopics:

Lobes of the cerebrum

Structures of the diencephalon

The cerebellum

Structures of the brainstem

Sample Prompt: Teach me about the lobes of the brain

The was a more variable approach where course topics were given along with sample prompts. The goal was to provide more flexibility in learning by allowing students to explore concepts if they wish to do so.

2025: Presently in development at the writing of this chapter. The new course chatbot will incorporate a module approach. The chatbot will begin by asking which module the student is working on and then present a list of topics they need to cover. Students can then choose the topic and the chatbot will work to teach the content with periodic knowledge checks. The chatbot will be designed to provide short chunks of content followed by questions and pauses for reflective writing or verbal responses. The chatbot will also provide feedback regarding understanding.

This will require a good deal of documentation that І will need to develop which will include the learning objectives for each module, important content points such as lists of anatomical structures and descriptions of processes. The goal is to provide a teaching session for students to complete before attending class so that the information can be applied to more advanced teaching cases or reinforced by completing labs.

Each module will require a document that will be uploaded to the chatbot. The chatbot’s master prompt will direct it to pull information from each document along with the course content and develop a personalized teaching session. Students will then copy and paste the entire AI interaction into assignments.

І have learned a good deal about student AI use by reading their AI interactions in assignments. As with many assessments, there is a range of abilities in prompting AI. Some students excel at learning by writing detailed prompts for their interactions and others do a minimal job of just enough to complete the assignment. A small number of students prefer not to use AI, which is allowed in my courses, and some do an excellent job of completing assignments without AI. І can work with some of the weaker students to improve their interactions and І hope to include some sharing sessions during class next semester so that students can learn from each other.

The in-class sessions include a good deal of small group discussions and activities, so there is opportunity for students to learn from each other and І have observed that some take advantage of this.

My other goal for this summer is to learn more about AI agents. AI agents go beyond chatbots in that they can access your computer to perform tasks whereby chatbots just provide information in the form of text, images, audio, or videos. For example, let’s say you wanted to plan a vacation. A chatbot could access the web and provide information about attractions, hotels, and transportation. An AI agent could go further and not only plan the vacation but also complete the hotel and airline bookings.

І don’t have a good use-case for an agent at present and І wouldn’t want it to access my personal or college emails, but І see agentic AI as a large part of the future of AI, so І want to explore this technology.

 

Summary Points 2024-2025

Here are some takeaways from the 2024-2025 semesters:

• Student use of AI is increasing dramatically

• It is important to openly discuss AI use frequently in class and create a safe environment for AI use

• Student AI prompting is improving

• It is nearly impossible to detect AI writing

• AI detectors are becoming more inconsistent

• It is vitally important to teach students how to use AI as a learning tool

• Educators should model ethical AI use and disclose their use of AI

• Use AI in class and help students analyze AI content

• Explore the use of AI in future careers

• Continually remind students why they need to learn course content

• Develop custom AI tools (chatbots) to help students learn course content and update these based on how students are using them

• Observe the use of AI in your classroom

• Evaluate student AI interactions (prompting, knowledge checks)

• Use AI to reach higher levels of learning including analysis and application

• Do not use AI to create “Pre-AI” activities and assessments but “stretch” your thinking to develop new types of activities and assessments

• Evaluate your course content to focus on the most important items and let students use AI to determine what these are

 

Overall, І am looking forward to the upcoming Fall 2025 semester and continuing my AI integration strategies. Most of all, І look forward to learning more about how my students are using AI and their concerns about the roles of AI in future careers.