Teachers and schools are questioning if they should ban generative AI technologies for students and contemplating how to execute such a ban if implemented. Ultimately, this is a waste of time. Generative AI shouldn’t be viewed as a threat, but rather processed into the fold of education with the proper use of the technology being clearly defined and explained to students.
What is Generative AI
AI, or Artificial Intelligence, is the use of computational systems to mimic human intelligence. When talking about AI, we’re mostly talking about combining the technologies of machine learning (ML), natural language processing (NLP), and some type of statistics or predictive analysis to determine confidence in the answer being determined.
While AI can perform a wide variety of tasks, Generative AI is a type of AI that can synthesize output using deep neural networks—machine learning designed to mimic the human brain itself—to process large volumes of input on the subject and generate new output that is similar in character to the original inputs. Whether viewed as mere mimicry or even as akin to the output process of learning, these newly produced outputs are increasingly realistic and uniquely formed, whether written text, images, music, spoken word, or even video.
Where can Generative AI be found?
The two primary providers of Generative AI technology, under the hood, are OpenAI and Google, with the former being the leader in the space. OpenAI tools you may have heard of include GPT-3 (and GPT-4 coming soon), Grover, and the now-famous ChatGPT which deploys the aforementioned GPT technology within a conversational interface. Google’s AI initiatives are best known through their LaMDA products, such as the upcoming BARD search updates. These tools, and independent AI models, are now built into software of almost every category or type.
Generative AI is most likely to be used by students when it has reached mainstream status, however. This means that the OpenAI tools are by far the most likely tools to be used by students for now. Written works created with ChatGPT's free tier are the most likely at the moment. But early adopters who have relied on generative AI for the longest are likely using third-party tools that give them access to GPT-3 or other similar AI models for writing text and copy: Jasper, Copysmith, Rytr, Frase, and the countless others you can find online.
With Microsoft investing a cumulative $15 billion in OpenAI to date, even the Bing search engine now has chat-style generative AI built in while Google still languishes behind. However, the use of ChatGPT within Bing Chat is still in an early preview as this post is being written, so it will be some time before it's adopted by anyone beyond the most stalwart of Microsoft fans. Bing distinctively bases all answers on web search results and cites each piece of data provided. This can be extremely useful for conducting research, as will be discussed next.
How Students Should Use Generative AI in Education
In my view, there is absolutely nothing wrong with the use of Generative AI by students who are old enough to be trained in both its usage and the processing of the resulting output. There are three primary ways that I openly support the use of Generative AI within my own classroom: tertiary research, focused microlearning, and proofreading.
Wikipedia has long been debated in academic circles. Historically, the question has been whether Wikipedia is a valid source of research. I've always found this quite concerning as it demonstrates that the detractors are insufficiently versed on research. Wikipedia, to continue this tangent, is a perfectly valid source of tertiary research. Tertiary research is research that is used to identify additional information or even sources for secondary research. So long as students follow the research trail to valid secondary sources, Wikipedia should be fair game although never cited directly.
Using Generative AI for research should be considered within the same framework. It's a valid tertiary source for data and information. As a tertiary source it shouldn't be cited directly in academic work. However, the output of Generative AI tools can lead students to valid sources of information that can assist with their research tasks, whether formal for peer reviewed work or informal for in class discussion.
Since Generative AI can create new output from a variety of vetted sources, it will become common to utilize it for focused microlearning. I tell my students that if they're having trouble understanding a concept through the various resources I provided to appeal to multiple intelligences... go ask AI. While currently viewed as heretical, this is actually similar to directing students to a textbook, online videos, or other sources that are known to be likely valid, but not necessarily perfectly aligned to the course. The purpose isn't so much to offload the teaching responsibility to technology, or a textbook, but rather to reinforce the existing teaching with an additional voice.
In my own undergraduate experience, I know that I spent many hours scouring textbooks that were only mildly aligned with coarse content for a better understanding of lecture topics and assignment requirements. By being able to ask a Generative AI specific questions about my understanding deficit, I would have been able to accelerate the learning process and spend much more time creating my own output and exploring through other means, such as kinesthetic, instead of spending the vast majority of my time on reading. As much of a fan of lecture and reading as I am (And I am!), even I have to admit that more verbal, visual, and kinesthetic learning would have been beneficial to my learning experience. By understanding this, I can try to offer that to my own students.
Many educators are already in the habit of suggesting tools for spelling and grammar check. While I grew up in a time where the use of spell check was debated, it's now virtually a non-issue to the point that we'd actively have to turn off spelling and grammar suggestions in most software, including the web browser. Gone are the days that spell check is a Microsoft Word-only feature.
So too will we see the day where tools such as Grammarly and others that are commonly suggested to students are integrated into our computing experience. My current web browser, Microsoft Edge, already has predictive text built in on the desktop. While we're used to this technology from mobile devices in order to save thumb strokes on the screen, the technology applied to the desktop is much more advanced. It understands what I'm writing and can offer contextual suggestions for how I'm most likely to finish a sentence, including referencing my main topic, and not just generic grammar suggestions.
With tools such as Grammarly being focused on student use, and many Generative AI text writing tools having rewrite features, it's unlikely for any classroom to not experience these within the next few months to years. Even primary education will be heavily impacted by these tool sets just as we currently are with spell check.
How to Counter the use of Generative AI by Students
Many of the tools you're likely already using to identify plagiarism from sources on the web are beginning to build in detection methods for identifying whether a paper was written with Generative AI. Other tests exist to determine if images and other creative works were created by AI, but those are beyond my current scope. If you are looking to keep synthetically created work out of your classroom for the time being, your best bet is to rely on these tools and follow their adoption of anti-AI algorithms.
The tool available to me currently, TurnItIn.com, is leading the charge and publishes semi-frequent updates on their AI detection initiatives.
Benefits of Generative AI for Teachers
Whether you've decided to eliminate AI from all student work completely, or openly embrace it by offering best practices for students to follow, you should at least consider how Generative AI could impact your own workflow. As a teacher your discernment of the appropriate usecases for a technology may be beneficial while conducting this analysis. In order to better understand the benefits of utilizing AI—from the teacher's perspective within the classroom—I asked ChatGPT why an educator would want to use AI for their own purposes. Here's what it said:
Generative AI provides a major advantage for educators by allowing for automatic generation of material to meet the individual needs of each student. This technology can help tailor teaching content quickly to various learning styles and aptitudes, which makes teaching more effective and engaging. Generative AI also helps teachers save time by automating the creation of personalized lesson plans, activities, and materials in minutes. This technology also opens up new possibilities to create different types of educational experiences with simulations, games, virtual field trips, and more.
My Classroom Best Practices for AI
In case you’re curious about my own views on allowing students to utilize AI within the classroom (an undergrad digital marketing course), I’ve adopted the following best practices.
1) AI is a welcomed addition to your work efforts.
2) You are encouraged to interface with trusted AI products (ChatGPT, Bing, etc.) to learn more about our topics.
3) You are encouraged to have AI help you structure your writing (outlines), but you must manually ensure compliance with class requirements, which AI still has difficulty doing.
4) You are welcome to review the results that AI produces for our class writing prompts, but you must still manually write your own work—even if it’s a derivative of the AI’s output.
5) You may use spellcheck, grammar check, and grammar rewrite tools as much as necessary to best conform to U.S. English.
6) Feel free to ask AI to make your writing more concise, but I’d urge you to not trust it to make your writing longer… the results are often light on depth and negatively impact scores in the analysis category of the rubric.
Tools Discussed Within This Post
- OpenAI ChatGPT - https://openai.com/blog/chatgpt
- Google LaMDA - https://blog.google/technology/ai/lamda/
- OpenAI GPT-3 - https://openai.com/blog/openai-api/
- Jasper - https://jasper.ai
- Copysmith - https://copysmith.ai
- Rytr.me - https://rytr.me
- Frase.io - https://frase.io
- Microsoft Bing Chat - https://bing.com/chat
- Grammerly - https://grammerly.com
- TurnItIn.com - https://turnitin.com