Co-Creational Teaching of Natural Language Processing

John P. McCraeACL 2024

TL;DR

This paper introduces "co-creational teaching," a methodology that transforms NLP education from traditional lectures into an active partnership between students and instructors. By co-designing curricula, engaging in peer teaching, and collaborating on projects, students gain deeper understanding and ownership, addressing the fast-paced and dynamic nature of NLP.

What Problem Does This Paper Address?

Traditional lecture-based teaching methods often fall short in delivering engaging and effective learning experiences, especially in rapidly evolving fields like Natural Language Processing (NLP). Students increasingly demand exposure to cutting-edge technologies, creating a challenge for educators to balance current trends with foundational skills. This paper addresses the gap between traditional pedagogical approaches and the need for dynamic, student-centered learning in NLP, aiming to improve student outcomes and engagement.

Methods

The paper proposes a "co-creational teaching" methodology built on three pillars:

  • Curriculum Co-design: This involves students and teachers collaboratively shaping the course syllabus. Methods include pre-course surveys, brainstorming sessions, and open voting on topics. The approach is inspired by "backwards design," ensuring learning outcomes align with student needs and interests.
  • Peer Teaching: Shifting from a lecturer-centric model, this approach leverages resources like YouTube videos and student-led instruction. The "jigsaw technique" is highlighted, where students become experts on specific topics and teach them to their peers, fostering deeper engagement and ownership.
  • Group Work: This pillar emphasizes collaborative learning, moving beyond traditional assignments. The paper suggests fostering a culture of controversy theory to encourage refined conclusions and structuring group work with clear roles and joint rewards to mitigate social loafing. The concept of the "Zone of Proximal Facilitation" is introduced, emphasizing the teacher's role in orchestrating tasks effectively.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Student Engagement is Key: The paper strongly advocates for active learning and student participation, arguing it leads to better knowledge retention and deeper comprehension.
  • NLP is Ripe for Co-creation: The fast-paced, interdisciplinary, and media-visible nature of NLP makes it an ideal subject for co-creational teaching, as students often arrive with prior knowledge and strong interest.
  • Beyond Lectures: Traditional lectures are deemed less effective, and the paper champions methods that empower students to become "their own teachers."
  • The Ladder of Student Participation: The paper references a model illustrating a transition from dictated curricula to full student control, with co-creation residing at the higher rungs of this ladder.
  • Addressing Generative AI: The rise of generative AI presents challenges for academic integrity, particularly in NLP. Co-creational methods, by focusing on in-class activities and collaborative design, can help shift assessment away from traditional essays and towards more verifiable, hands-on learning.
  • Neurodiversity Considerations: The paper acknowledges that while co-creational methods can be beneficial for students with ASD and other neurodivergent traits, careful scaffolding and clear guidance are crucial to accommodate varying interaction styles and learning preferences.

Key Results

The paper details an application of co-creational teaching in an MSc course on "Advanced Topics in Natural Language Processing" at the University of Galway.

  • Positive Feedback: The course received strong positive feedback regarding student engagement and overall outcomes, with students performing well on the written exam.
  • Curriculum Evolution: The syllabus was dynamically developed through student feedback and voting, resulting in a curriculum that was more prescribed than fully student-led but still significantly different from a traditional lecturer-designed course.
  • Project-Based Learning: An open project component allowed students to explore topics of their choice, fostering ownership.
  • Challenges Identified: Despite successes, issues arose, including students not consistently watching pre-lecture videos and some students engaging with peer assessment infrequently. The paper notes that some students chose to work alone, leading to less engagement.

Limitations and Future Work

  • Theoretical Focus: The paper acknowledges that its primary contribution is theoretical, with only one empirical evaluation in a single setting. Further quantitative analysis and application in diverse settings are needed.
  • Scheduling and Planning: Co-creational methods require more spontaneous scheduling, which can clash with institutional timetabling and make advanced planning, especially for exams, more difficult.
  • Lecturer Workload: While less content preparation might be needed in a traditional sense, preparing content anew each year and adapting to evolving topics can increase lecturer workload.
  • Student Engagement Variability: The paper notes a risk that not all students will be engaged, potentially leading to poorer outcomes for less engaged individuals.
  • Future Directions: The authors plan to focus on improving peer instruction and applying more of the discussed techniques in future iterations of the course. They also suggest adapting these methods for students with neurodiversity, emphasizing clear guidance and flexible interaction methods.

Summary updated

BibTeX
@inproceedings{mccrae2024co,
  author = {John P. McCrae},
  title = {{Co-Creational Teaching of Natural Language Processing}},
  booktitle = {Annual Meeting of the Association for Computational Linguistics},
  year = {2024},
}
Figure