Leveraging AI

Data Analytics in an Academic Setting

This hands-on workshop will walk through some basics of using AI tools such as ChatGPT to quickly analyze large-scale text responses such as open questions in student response forms. We’ll explore sentiment analysis to identify emotional trends, trend detection to track changes, and general feedback analysis to gather actionable insights from student feedback. We’ll also discuss the importance of protecting personal information and the limitations of data analysis with the tools provided.

Agenda:

Definitions
Possibilities
- - Sentiment
  - Trend
  - Feedback
Limitations
- - Getting Your Data In
  - Verifying Your Results
Sample Data & Sample Questions

Linguistic Analysis

within large pools of text, analyzing words, sentences, and meaning for the purpose of identifying patterns and insights.

Aggregated dataset

a compilation of data sourced from multiple records or locations. This dataset is systematically grouped or summarized based on shared criteria to enhance its usability or scope. It is often stripped for broader use.

Stripping Data

with respect to PII (Personally Identifiable Information), involves carefully removing or obscuring any data elements that could be used to identify an individual. Beyond names and addresses, this may also involve underrepresented ethnicities, distinct gender expressions, or unique characteristics.

Loading and Using Your Data

Input Characterizations

PII: always strip your data FIRST.
Tell the AI what you're loading.
Do an Input Analysis SECOND.
- Number the responses.
- Ask the tool to count how many responses it registered.
- Close enough?

Important Limitations

It’s not a paid data analysis tool. You will have to do some work.
Prompt additions for accuracy:
1. - “Limiting yourself to ONLY the dataset I just provided,”
Double check the answers:
1. - Give 10 examples of responses that line up with “Answer A”
  - Count the total number of responses that line up with “Answer A”
  - Estimate the proportion of responses that line up with “Answer A”

Types of Analysis

sentiment analysis

What emotional tone is present in this data?

Categorize each response as positive, negative, or neutral and provide a summary of the sentiment trends observed.
Identify the sentiment in response to “Topic B.”
Categorize and identify potential underlying reasons for these sentiments.

Potential Prompts:

Limiting yourself to ONLY the dataset I just provided, What is the overall sentiment of the student responses—positive, negative, or neutral?
Limiting yourself to ONLY the dataset I just provided, Which specific courses or departments received the most positive feedback?
Limiting yourself to ONLY the dataset I just provided, Are there any recurring negative sentiments regarding particular aspects of the student experience, such as course difficulty, instructor support, or facilities? (Another way to ask this is “common concerns or issues”)
Limiting yourself to ONLY the dataset I just provided, How do students feel about the online learning experience compared to in-person classes?
Limiting yourself to ONLY the dataset I just provided, Are there noticeable changes in sentiment over time, such as improvements or declines in student satisfaction? (This only works with dated responses)

Trend Detection

Given two or more sets of data, how do they differ?

Analyze the enrollment data for the past five years; identify and report on any emerging trends in student types.
Identify trends across geographic regions.
Highlight courses gaining popularity and predict upcoming trends for next year.

Potential Prompts:

Limiting yourself to ONLY the dataset I just provided, What are the most common themes or topics mentioned in the student feedback?
Limiting yourself to ONLY the dataset I just provided, Are there any emerging trends in student preferences for course delivery methods (e.g., online, hybrid, in-person)?
Limiting yourself to ONLY the dataset I just provided, How have student perceptions of the college’s support services (e.g., tutoring, library, counseling) changed over recent semesters?
Limiting yourself to ONLY the dataset I just provided, Which areas of the student experience are frequently highlighted as strengths or weaknesses?
Limiting yourself to ONLY the dataset I just provided, Are there any patterns in feedback from different demographic groups (e.g., age, major, year of study)?

Feedback Analysis

Extracting actionable insights from the data.

Identify common themes and concerns raised by students and quantify how many had positive versus negative experiences.
Categorize and summarize feedback into change requests, issues reported, and general feedback.
Extract key themes related to learning environment, college efficacy, and student sense of belonging. Provide insights into strengths and weaknesses.

Potential Prompts:

Limiting yourself to ONLY the dataset I just provided, What specific suggestions do students provide for improving course content and delivery?
Limiting yourself to ONLY the dataset I just provided, Are there recurring issues or challenges mentioned by students that need addressing?
Limiting yourself to ONLY the dataset I just provided, What positive feedback do students give about their instructors and teaching methods?
Limiting yourself to ONLY the dataset I just provided, How do students rate the effectiveness of college resources such as the library, career services, and study spaces?
Limiting yourself to ONLY the dataset I just provided, What are the key areas where students feel the college is excelling or falling short in supporting their academic and personal development?

A sample file is available for GCC employees who would like to try out these prompts. Email Loy to request it.

Google Sites

Report abuse