I have weekly conversations with students and professors about using artificial intelligence (AI) in their teaching. What I call "pairing" is one of the fundamental teaching strategies I advocate for. This method is part of a family of approaches with a long history of success in pedagogy, and I have found it beneficial in my teaching. This is my description of it.
If there is a chance that students will purposefully or unintentionally exploit AI to complete one of your assignments, you should find strategies to discourage them from doing so. You should search for other tasks that may be used with this susceptible assignment and are intended to encourage students to finish it without abusing artificial intelligence.
Consider an assignment that is simple to do using ChatGPT and doesn't require the student to study or put in effort to fulfill its goals. Let's assume they can enter your prompt and rubric into ChatGPT, upload a few readings, and presto! They will have an essay that meets the rubric's requirements and is ready for submission.
Professors have adopted the tactic of assigning this kind of essay while issuing dire cautions on the misuse of AI. They then summon the student into their office and ask them whether they offloaded their work onto the LLM when they think the student submitted work that ChatGPT generated.
This check or layered method is incorporated into the assignment design via pairing.Â
By employing this method, the professor assigns the essay along with an oral exam or tutorial, which, like the professor-student meeting described above, enables the professor to assess the degree to which the student has understood the material the essay intended to teach.
The student is motivated to finish the essay in a way that satisfies the learning objectives of the essay because they are aware of the connection between their writing and the oral test, even if this still requires some use of artificial intelligence. To properly incentivize students, the oral test should be given more weight than the essay.
Extrapolating From Pairing
Cases unrelated to academic dishonesty can also benefit from the pairing.
Suppose a professor is worried about "hallucinations" or other subtle errors caused by AI in student submissions. In that case, the assignment should be assessed to encourage students to find and fix the mistakes before turning in their work.
In this situation, pairing provides the lecturer with a potent means of aligning incentives. A second assignment might be given together with the first one, which is specifically made to assess how much the students' AI-generated work contains faults. Like the professor-student meeting, this second task should be precisely the exam one would do to identify the student's mistakes.
For instance, it can entail selecting sources randomly to ensure they back up the students' assertions about their support for the facts. A paired second assignment, appropriately weighted, places the necessary emphasis on the student avoiding AI misuse or errors rather than treating it as a component of the first assignment. However, such a sampling feature might also be incorporated into the grading of the first assignment. (To take advantage of peer review and align incentives, you may give a peer the second task.)
If a university student is correct that "only a small group of students choose to attend college primarily out of love for intellectual discovery," then AI generally puts pressure on teachers to better align the incentives their students experience. (I think the student's motivational narrative is far more complex than his own, but he's partially incorrect.)
Alternative Techniques
By improving incentive alignment, other tactics for preventing AI abuse center on incentives. I contend that there are, in general, six primary strategies to deter and stop students from abusing AI on a particular task:
1.        Encourage students to finish their assignments without abusing AI.
2. Â Â Â Â Â Â Â Demand that pupils finish the task without using artificial intelligence.
Given the growing strength of local LLMs and SLMs, AI may be readily accessed on devices with or without an internet connection, leaving two possibilities that do require a device. Provide a handwritten and an oral version of the assignment for the students to complete.
3. Â Â Â Â Â Â Â Give students access to AI to finish a (more) AI-immune version of the assignment.
Although creating an assignment that is more AI-immune complicated and
always-changing process, there are two main types of options: Creating an
assignment that is AI-immune because
4. To encourage students to meet the learning objectives in both situations, pair the
assignment with another one they must do in an AI-free environment.
You can think about partnering differently, like how you feel about students' capacity to rely on the knowledge of their classmates in the dining hall and their dorms. They can certainly ask their astute friend for advice on how to draft an essay or solve an issue. Still, they must use those abilities and knowledge internally (rather than just memorizing it) when attending class if its format and content are the first steps. Â
5.       Do nothing
6. Combine any of the above
           In summary, Pairing is a collaborative teaching strategy rooted in the principles of cooperative learning. In this method, two students are intentionally grouped to work together on a task, project, or discussion. This method draws from Vygotsky's social constructivist theory, which emphasizes the role of social interaction in cognitive development and has a long history of efficacy in fostering academic and people skills.
In pairing, learners are often chosen to complement each other based on their skills, knowledge, or perspectives, creating an environment where mutual support and peer-to-peer learning can flourish. This approach operates on the premise that dialogue and collaboration enable deeper processing of information, scaffold learning experiences, and promote critical thinking. Students take on dual roles as learners and teachers when paired effectively, contributing to reciprocal learning benefits.
The key features of pairing include:
Active Engagement: Pairing encourages all participants to engage with the material actively, reducing the risk of passivity often associated with lecture-based instruction.
Differentiated Support: Teachers can pair students strategically to provide differentiated support. This allows stronger students to reinforce their understanding by explaining concepts while allowing peers to gain experience in a less intimidating, one-on-one setting.
Social and Emotional Learning: The method inherently builds people's skills, such as communication, empathy, and collaboration, which are critical for lifelong learning.
Pairing is versatile and adaptable to various instructional goals, making it applicable across disciplines and educational levels. Examples of its application include peer tutoring, think-pair-share activities, paired reading, and partner problem-solving. The success of this method is contingent on thoughtful pairing, clear task structures, and ongoing teacher facilitation to ensure balance and equitable participation.
In conclusion, pairing is a time-tested and research-supported pedagogical strategy that enriches the learning experience by leveraging the collaborative power of peer interaction. Its flexibility and effectiveness in addressing cognitive and social aspects of learning make it a cornerstone of successful teaching practices.
References:
Acemoglu, D., & Restrepo, P. (2020). Robots and jobs: Evidence from US labor markets. Journal of Political Economy, 128(6), 2188–2244. https://doi.org/10.1086/705716
Brynjolfsson, E., Rock, D., & Syverson, C. (2019). Artificial intelligence and the modern productivity paradox: A clash of expectations and statistics. NBER Working Paper Series, 24001. https://doi.org/10.3386/w24001
Felten, E., Raj, M., & Seamans, R. (2019). A method to link advances in artificial intelligence to occupational abilities. AEA Papers and Proceedings, 109, 58–63. https://doi.org/10.1257/pandp.20191016
Comments