This post was originally published on this site.
A generative artificial intelligence tutor in Macmillan Learning’s software helps address student questions as they complete assignments.
Macmillan Learning
Over the past two years, use cases for generative artificial intelligence in higher education have grown, offering opportunities for experiential learning, building course materials, academic advising and research to support students.
Students have been early adopters of generative AI tools for their coursework, turning to chat bots for help with essay writing or answering tough homework questions. Increasingly, staff and faculty members have joined the fray, using generative AI for design, content and assessment in the classroom, as well as for writing emails or fielding questions from students outside it.
AI-powered chat bots have grown in popularity, with an Inside Higher Ed survey this fall showing 50Â percent of chief technology officers are building their own chat bots and assistants.
Most Popular
New data from Macmillan Learning finds AI tutors can assist in student learning and skill-building, as well as increase learner confidence to ask questions and dig deeper into materials.
The background: Macmillan collected student usage data as they engaged with the tech company’s AI Tutor, a homework support tool that is embedded into Macmillan Learning’s digital platform, Achieve. As the name suggests, Macmillan’s AI Tutor is a generative AI–powered homework support that provides guidance and answers course-related questions around the clock. At present, the tool is available for STEM and economics courses.
“The tutor is like an extension of an instructor or teaching assistant that guides them without judgment, or the stigma of asking basic questions, whenever they need support,” according to a December press release.
Similar to other teaching tools, AI Tutor won’t give the student the exact answer to their query but will direct them in ways that boost students’ self-efficacy and problem-solving skills using Socratic questioning.
Macmillan analyzed more than two million messages from 8,000 students in over 80 courses from fall 2023 to spring 2024.
Researchers also gathered feedback using student and instructor surveys. The student surveys were fielded in fall 2024 at nine institutions as two-week regular check-ins, so student response rate varies by question.
Students say: The average user engaged with AI tutor for 6.3 minutes, with some sessions lasting up to 30 minutes.
The AI Tutor, when toggled on by the professor, is embedded into STEM courses to support student learning.
Macmillan Learning
AI-powered homework tools have been touted as supporting students outside of the classroom at hours when human instructors may be unavailable, which Macmillan’s data affirmed: Students use the tool most from 6 p.m. to 3 a.m. on weeknights.
Feedback also showed students felt comfortable engaging with the tool in ways that they wouldn’t with a faculty member. One student said they felt they would be annoying a real person with their questions compared to the chat bot, which was patient and helpful. Another student didn’t bother with honorifics, writing to the AI tutor, “bae i dont get this.”
In a survey of instructors (n=30), 41 percent of respondents say the tool improved students’ confidence or exam performance, and nearly half say students seek help from them less with the use of the AI tutor. Professors also say students are more prepared for class and have a better understanding of content, which allows them to cover more advanced ideas and engage in richer discussions.
Within Achieve, faculty have the option to turn the AI Tutor on or off for the entire class, allowing the instructor to dictate the course’s AI use policy.
After using the tool, 44 percent of students (435 students, four surveys, 207 responses) reported increased confidence in their problem-solving abilities and, over time, a quarter of students say they use it less because they’ve honed their own skills.
Many in higher education worry students will use generative AI dishonestly or as a crutch to impede learning, but students say they’re self-regulating when and how they use the tool. Results from one survey (n=582) show 67 percent of students say they only use the AI Tutor when necessary and 29 percent are using it more than needed “sometimes.”
A majority of learners said they’d be disappointed if the tool was no longer available, but students indicated concerns about equity, with some respondents indicating AI tutors should be made available equally to all learners in the course.
How is your institution measuring the effectiveness of AI-powered tools in the classroom? Tell us more here.