AI in education- is it the schools that are cheating?
"Quis custodiet ipsos custodes?"
This Latin phrase translates to "Who watches the Watchmen" - questioning oversight for those in power.
When it comes to AI in Education - should it be the students that are worried about the honesty of the Schools?
the fear of student academic dishonesty with ai
Academic Dishonesty has long been a concern across education. With the most reputable schools being the most concerned about ensuring that students are providing their own work for assessment. This concern comes from a desire for authentic learning and honest work.
The general availability of AI tools to the public has created a range of thoughts on how AI will change how students use - or mis-use - the tools for their work. This has led to an entire industry of AI detectors and anti-AI sentiments. From banning to detecting to supervision - AI has been the latest technology to disrupt the classroom and draw the attention of those fearing ill-use of the tool.
While Schools struggle with the question of how AI will disrupt their teaching and ability to ensure authentic learning - it is also fair for students to ask the exact same questions.
Not all schools or online course providers are reputable
With education being a multi-billion dollar industry, there will be those with less than ideal morals and ethics. The diploma-mill schools, pay-to-PhD schemes, and academic dishonesty by those in charge of education have all created an environment where diligence is required by students.
There will be those that seek out these types of business’s, but I remain optimistic that that is the exception and not the norm. For most students, education represents an investment that requires sacrifice, family support, savings, delayed opportunities. Shortcuts may be tempting - especially when combined with false promises.
The cost of education for the students has only risen and they must ensure they are not delivered a sub-par, generic, and unaligned educational experience. Sadly, there are even cases of once reputable schools finding themselves in pay-to-admit schemes and lax standards.
If Schools are using ai to create learning materials - are they disclosing it?
Economic conditions demand faster course development and portfolios that adapt to market demands. Schools are always trying to do more with less. There is an argument to be made around the role of “additive” thinking vs. “multiplicative” thinking, but that’s a post for another time.
This pressure to develop and release new courses must not become a temptation for Schools to cut corners and fall into the trap of only using automated tools such as AI to create or heavily influence the process of learning creation. If this becomes the norm, then we will simply have generic, information exchanges as opposed to true learning and teaching.
Courses are a tool for learning - Schools must first and foremost focus on the learning process and then determine mechanisms to make that learning occur. (Again, more on that is some subsequent BLOG Posts I’ll create).
At the very least, I believe it is important for schools to disclose when a course is AI-generated or assisted. This is Academic Integrity. In the same way there is an expectation on students to cite their work - schools should cite theirs.
Traditional education, open, and online
AI is everywhere in education. This means that we will see it being used across all types of course creation no matter the type of courses being viewed. A quick search on “How to create a course quickly” will yield all types of AI tools that can generate content which, often, is then promoted as a way to earn money online.
This isn’t untrue, but it’s also not completely honest. To either the creator of the content and especially the students. As you will see shortly, creating a course is easy. Creating effective learning is a skill and not going to be replaced by AI any time soon in my opinion.
While the AI tools are often promoted as “build and sell your own course” tools - their use has permeated all layers of education. Traditionally, we have has a perception of education being “tiered” - Ivy League to Community Colleges to Trade Schools to Specialty Schools to Online Schools to Some Guy on YouTube classes. Even I have an online course around the Microsoft Whiteboard! (Which, full-disclosure was not AI generated).
At one time, these divisions were based on reputation and perception. We’ve seen a huge backlash against tradition with the “drop out of college” and “teach yourself” movements. These may have grown out of economic frustration at tuition costs, or the perception that the quality/value of non-traditional training has increased while traditional education has stagnated, become outdated, and does not reflect the world and deliver the value it once did.
There are more options than ever for potential learners and the demand for specific methods of education are less important than demonstrable skills: certifications, portfolios, and the ability to show your work. This means that “Performance beats Paper” in the minds of many. Societal attitudes are changing on the value of a specific educational output such as a Degree.
The experiment
To demonstrate just how a potential course creator (or School) could “cheat” using AI to develop a course, I thought that I would give it a try.
Most course outlines are somewhat of a template. You have a course description, some desired length for the training to occur, modules, and activities. For each module, you will have some learning outcomes and objectives plus some readings/review material and a lab or exercise. Often leading to an assessment of some type.
This applies to almost all learning types: in-class, remote, synchronous, asynchronous, formal, or commercial. You can see this structure from the largest University undergrad classes to a typical online platform.
The goal of my experiment was to see what AI could do at this initial stage of course creation and definition.
Creating the prompt
For the first phase of this experiment, I decided to go with my ChatGPT4 account and run the prompt there. This ended-up being sufficient for my initial test - as you will see in the upcoming results.
Here is the prompt:
I would like help creating a 15 week course that prepares students to learn computer networking. This course should have 12 modules. Each module should have a module description, 2-3 learning outcomes, and 5-7 learning objectives that relate to the learning outcomes.
There should be 2-4 recommended readings with links to the readings for each module as well as a suggested learning activity.
Can you create this for me?
This prompt was a basic course outline and certainly could have more elements. ChatGPT is limited in any one response, so I wanted to ensure that it was specific enough, but not asking for too much at once.
When working with AI prompts, I’ve found an interactive approach is always best.
Creating the course - results
The results were… surprising. At least to someone that has been used to creating course outlines more manually.
The results for this course were quite extensive - although the complete result did require me to occasionally prompt ChatGPT to continue - once every 4 or so modules. The results also had hyper-linked readings - the PDF above stripped those from being active links.
If you are considering this experiment, I encourage you to just replace the term “Computer Networking” in my prompt with a Domain of knowledge you are familiar with. This will allow you to understand the power (and limitations) of AI used in this manner.
In my case, the outline did look like a generic, average Computer Networking course, but as I have and do teach this subject, I immediately noticed flaws. My experience allowed me to see obsolete subject matter, poor ordering choices, and deficiencies around context of the skills being sought (notice I didn’t say taught).
My next question was: “how does this outline compare to actual courses being offered?”
Comparing the results to online options
In order to gain some insight into how this AI-generated outline stacked-up to courses offered online, I visited several popular online course platforms.
While the subject areas were very similar to what both the AI-generated and traditional school Course Outlines had (even down to the order of subjects) - there was one main difference between the AI-built outline and the online courses: videos. The online courses not only has the structure of the AI Outline - but also a list of videos that you would be watching.
Comparing the results to in-person schools
The next comparison I did was to get course outlines for some traditional colleges for a Computer Networking course from across several regions. In this case, I won’t list specific Schools I visited, but you can certainly do some searching for ones in your areas and access course outlines.
This was, again, a very close match to what ChatGPT had generated. Which, was slightly concerning to me. My search was not exhaustive, but seeing the level of matching made me question what, if anything, these institutions were offering that couldn’t be both organized and fulfilled by AI and its synthesis models.
Maybe a basic Computer Networking course is the same content mapping no matter the format or place where it comes from? This may not be a good look for places trying to convince you to attend their offering over online over just using AI to learn. More on that later.
When differences did occur, what were they?
While online courses had mostly videos to demonstrate and explain the concepts, it was traditional schools that had the potential to offer more - but just the potential. Not all traditional schools seem to tap into broader delivery and teaching elements.
I didn’t explore the course delivery methods of all the courses I was reviewing as I was more interested in just the initial outlining phase to build the Course roadmaps. However, the learner will first interact with any Course Outline - which becomes important. In some cases, the outline is the “contract” with the learner.
What makes the AI-Generated outline even more interesting is the argument that once the structure is generated by AI that videos or other learning resources can easily be found for each element. This is one of the reasons I included the request for “Selected Readings” in my ChatGPT prompt - to see how well the AI could find related content. The prompt could be modified to ask for readings, videos, lab guides or more.
For traditional schools, the delivery mechanism is less clear in their online descriptions of the courses. Effectively their website will suggest “register and attend class”. Presumably you learn about what happens next when you attend the class - but maybe they should be more upfront about the specific process they will engage in to help you learn.
Students should ask the school how they create courses and how their courses deliver value
It’s impressive how new generations are savvy about value for their time and investments. They have grown up in an environment where they are able to access information and avoid gate-keepers that attempt withhold information from them.
It has become easy for them to just copy a Course Description, use AI to generate the outline, and then find free or low-cost ways to acquire the knowledge they seek.(check-out my YouTube Channel!).
This means that if a course provider wants to attract these learners into their programs, they will need to have a clear value-proposition that goes beyond information exchange and into a transformative experience - with clear evidence and processes on how this transformation will occur.
Smart students will ask what the key differences are between their choices - and, as we see the evolution of educational expectations, it won’t be the paper at the end. It will be the knowledge acquired. Which, personally, makes me very excited about the future.
Schools should promote differentiators
If students are asking - schools should be answering.
Some of the areas I can see traditional schools providing key added value over AI-Generated content is around the following:
Resources such as equipment and labs
Mentoring and Guidance from Instructors (as opposed to just information delivery)
Cohort activities for teamwork and collaborative opportunities
resources and labs
This is a very strong offer for many technical schools. They have industrial equipment, kitchens, printing presses, studio equipment, computer networking labs and more. This gives students a strong incentive to select these over AI-generated courses that can only point to online resources such as videos and possible simulators.
As other areas such as VR/AR and remote labs accessed through the Internet mature - some of these resource advantages may become threatened. However, the smart schools will use newer technologies to enhance rather than replace. A VR kitchen isn’t a great way to make a real pie.
Mentoring and Guidance
Smaller class sizes with more interactions with Faculty are an area where schools can also differentiate themselves from massive online classes (or even the 400-person lecture auditorium). There is still an attraction to having a real person available during a learning experience to guide and manage learning.
While a University once was able to promote their classes to students based upon attracting especially prominent Professors - most students would not directly interact with those Professors until at least the Graduate or Post-Graduate courses. An undergrad likely has more access to the Professor through YouTube videos or published papers than in a lecture hall.
By building programs that are collections of courses built as a reenforcing system of learning, Instructor-led training can be more than just information exchange.
Cohort and team activities
Learning with others and building projects together is something that schools absolutely have an advantage over AI/online content.
This is not a complete list, of course. For some students the “College Experience” and less tangible aspects will play a role. However, at the cost of the “College Experience” many students may opt to learn online and have a “Life Experience” through travel or other mechanisms.
My conclusions Your own experiments
Using AI to create a Course Outline reminded me of and taught me a few things:
First: AI does a great job of creating a generic course outline. In many cases, the outline was no different or even better than the outlines I found online and from schools. Perhaps the ones I found were the input for the AI’s work or perhaps AI was used by the schools I looked at? No matter how each outline was created - the AI-generated outlines were a very close match.
Second: AI wasn’t flawless (and neither were the schools). This particular subject is one that I’m familiar with (Computer Networking) so I immediately spotted flaws in the outlines.
Third: Schools must adapt and add context over content. Information is everywhere, cheap and available. Knowing how to organize, construct and use that information in pursuit of knowledge does still require teaching and teachers (which, I’m grateful for!). Teachers are, in my opinon, much better and gauging and adjusting learning for classes and individuals than an AI. An AI is good at content synthesis - but content synthesis is not learning by itself.
There were more take-aways for me, but it’s your turn!
I hope you will take the prompt I’ve provided, change the term “Computer Networking” to an area that you have Domain knowledge and see the results.
Comments