Creating Tailored Chatbots GPT with Cogniti: Empowering Teachers with Insightful Feedback
- Simon A
- Jun 23, 2024
- 4 min read
In the evolving landscape of education, leveraging technology to enhance learning experiences is paramount. One such innovative tool, Cogniti, developed at the University of Sydney by Professor Danny Liu, utilises Microsoft Azure GPT-4 to bring cutting-edge AI capabilities into the classroom for teachers. Cogniti allows teachers to create their own tailored chatbots and observe the exchanges between students and the AI. This setup provides educators with raw data that can be transformed into insightful reports, helping to better understand and support student learning.
I have been using Cogniti with two Year 11 Business Studies classes (approximately 50 students) for 1-2 months in NSW, Australia. Importantly, the students were not forced to use Cogniti; much of the data comes from organic usage, highlighting genuine engagement. I can gain profound insights into my students' learning behaviours, needs, and areas of interest. The following sections explore how Cogniti is being utilised in these classes and how it enables teachers to better understand their students through data-driven insights. The raw data was used to create these reports with the help of Chat GPT and R.
Focus on Syllabus Areas
The frequency of questions by category over time provides a clear picture of the specific syllabus areas where students seek help. The graph reveals peaks in activity, particularly around critical study periods like the end of May and mid-June, indicating heightened student engagement likely driven by upcoming assessments or deadlines.
Basic Definitions: This category shows significant peaks, especially at the end of May and early June. Students frequently seek clarification of fundamental concepts, which suggests a need to solidify their understanding before exams. Teachers can use this insight to reinforce basic concepts during these periods.
Business Management and Finance: Consistent activity in these categories highlights their importance in the curriculum. Students regularly ask questions about management strategies, financial principles, and related topics, indicating these are core areas where they seek additional support.
External Influences, Marketing, and Human Resources: These categories also show regular engagement, reflecting a well-rounded interest in various business aspects. Questions in these areas suggest that students are keen to understand the broader context of business operations, including how external factors influence business decisions, marketing strategies, and human resource management.
Financial Ratios and Formulas: The engagement in this category indicates that students are working on understanding and applying financial metrics, which are crucial for analysing business performance. This can guide teachers to focus on these topics when preparing students for assessments.
Operations: Although this category shows lower frequency compared to others, it still indicates that students are interested in the practical aspects of business operations. Teachers might consider incorporating more real-life examples and case studies to enhance understanding in this area.
Case Studies: The relatively low number of questions related to case studies suggests that students might prioritise theoretical understanding over practical application. This insight can guide educators to integrate more case studies into the curriculum, ensuring students develop a well-rounded comprehension of business concepts.
By understanding the frequency of questions across these categories, teachers can better tailor their lesson plans to address areas where students need the most support, ensuring a comprehensive and targeted approach to education.
Sentiment Analysis of Student Questions
The sentiment analysis of student questions reveals interesting insights into the emotional tone of their inquiries. Most questions have sentiment scores clustered around zero, indicating a neutral sentiment. Neutral sentiment in this context means that the questions are primarily factual or informational, without strong emotional undertones. These could be straightforward inquiries about business concepts, definitions, or processes.
A smaller number of questions show slightly positive sentiment scores. Positive sentiment indicates that the questions might include expressions of curiosity, enthusiasm, or satisfaction. For example, a student expressing excitement about understanding a concept or asking how to apply a newly learned idea in a real-world scenario might exhibit a positive sentiment.
The prevalence of neutral sentiment underscores the academic and objective nature of the interactions between students and Cogniti. The occasional positive sentiment suggests moments where students feel engaged or pleased with their learning progress, offering valuable feedback for educators.
Hourly Interaction Patterns
Analyzing interactions by hour of the day uncovers fascinating patterns. The highest activity is observed in the early hours, from midnight to 2 AM, suggesting that students are utilizing Cogniti for late-night study sessions and last-minute revisions. Another peak in the late evening highlights the flexibility of Cogniti in accommodating diverse study schedules. While these patterns demonstrate the dedication and determination of students, they also underscore the importance of promoting healthy study habits. Teachers can use these insights to encourage students to manage their time effectively and avoid working excessively late, ensuring they get adequate rest and maintain a balanced study routine.
Depth of Engagement
The scatter plot showing the number of interactions versus average answer length per conversation highlights the variability in student engagement. While many students engage in short, concise exchanges, there are instances where prolonged interactions correlate with significantly longer responses. This indicates that while some students seek quick answers, others delve deeper into the subject matter, necessitating detailed explanations. Recognising this diversity in engagement helps teachers tailor their instructional approaches to meet the varied needs of their students.
Interaction Distribution
The distribution of the number of interactions per conversation reveals that most interactions are brief, with a significant number of conversations consisting of 1 to 5 exchanges. However, there is a notable tail of conversations extending beyond 10 interactions, signifying deeper engagement from a subset of students. This distribution helps teachers identify students who may need more attention and those who are independently navigating their learning journey.
Using Cogniti in the classroom provides teachers with a wealth of data-driven insights into their students' learning behaviors. By analyzing engagement patterns, sentiment, interaction times, and the depth of engagement, teachers can better understand and support their students. Additionally, Cogniti allows educators to view detailed data on individual student interactions, enabling a personalized approach to education. I have been fortunate to be able to test this out in my classroom, and it has significantly enhanced my ability to provide targeted feedback and support to my students. If you are interested in trialing Cogniti in your own classroom, please feel free to get in contact to explore the possibilities. This comprehensive understanding helps educators to create tailored feedback, develop targeted teaching strategies, and foster a more effective and responsive learning environment. Cogniti not only enhances the learning experience for students but also empowers teachers to become more attuned to their students' needs, ultimately leading to improved educational outcomes.
