Kids who use ChatGPT as a study assistant do worse on tests
A University of Pennsylvania study found high school students using ChatGPT performed worse on math tests, indicating that reliance on AI may hinder learning and problem-solving skills despite improved practice performance.
Read original article
A study conducted by researchers at the University of Pennsylvania revealed that high school students using ChatGPT as a study aid performed worse on math tests compared to those who did not use the AI tool. The experiment involved nearly 1,000 Turkish high school students, where those with access to ChatGPT solved more practice problems correctly but scored 17% lower on subsequent tests. A separate group using a modified version of ChatGPT, designed to provide hints rather than direct answers, showed improved practice problem performance but did not achieve better test scores than students who practiced without AI assistance. The researchers concluded that reliance on ChatGPT may inhibit learning, as students often sought direct answers instead of developing problem-solving skills. Additionally, ChatGPT's inaccuracies contributed to the issue, with the chatbot providing correct answers only half the time. The study highlighted a concerning trend of overconfidence among students, who believed they learned effectively despite poor test results. The findings suggest that while AI can assist in practice, it may not enhance actual learning outcomes, echoing historical concerns about technology's impact on education.
- Students using ChatGPT performed worse on math tests than those who did not.
- A modified AI tutor version improved practice performance but did not enhance test scores.
- Overreliance on AI tools may inhibit the development of problem-solving skills.
- ChatGPT's inaccuracies contributed to students' poor performance.
- Students exhibited overconfidence in their learning despite lower test results.
Related
AI can beat real university students in exams, study suggests
A study from the University of Reading reveals AI outperforms real students in exams. AI-generated answers scored higher, raising concerns about cheating. Researchers urge educators to address AI's impact on assessments.
How Good Is ChatGPT at Coding, Really?
A study in IEEE evaluated ChatGPT's coding performance, showing success rates from 0.66% to 89%. ChatGPT excelled in older tasks but struggled with newer challenges, highlighting strengths and vulnerabilities.
Can ChatGPT do data science?
A study led by Bhavya Chopra at Microsoft, with contributions from Ananya Singha and Sumit Gulwani, explored ChatGPT's challenges in data science tasks. Strategies included prompting techniques and leveraging domain expertise for better interactions.
Why AI is no substitute for human teachers
A study from the Wharton School found high school students using generative AI for math prep perform worse on exams, highlighting the need for guidance and the importance of human teachers.
Kids who use ChatGPT as a study assistant do worse on tests
A University of Pennsylvania study found that high school students using ChatGPT scored worse on tests, while a specialized AI tutor improved problem-solving but not test scores, highlighting potential learning inhibition.
11 commentsI have a feeling that is an unintentional #2 that dominates reality.
Related
AI can beat real university students in exams, study suggests
A study from the University of Reading reveals AI outperforms real students in exams. AI-generated answers scored higher, raising concerns about cheating. Researchers urge educators to address AI's impact on assessments.
How Good Is ChatGPT at Coding, Really?
A study in IEEE evaluated ChatGPT's coding performance, showing success rates from 0.66% to 89%. ChatGPT excelled in older tasks but struggled with newer challenges, highlighting strengths and vulnerabilities.
Can ChatGPT do data science?
A study led by Bhavya Chopra at Microsoft, with contributions from Ananya Singha and Sumit Gulwani, explored ChatGPT's challenges in data science tasks. Strategies included prompting techniques and leveraging domain expertise for better interactions.
Why AI is no substitute for human teachers
A study from the Wharton School found high school students using generative AI for math prep perform worse on exams, highlighting the need for guidance and the importance of human teachers.
Kids who use ChatGPT as a study assistant do worse on tests
A University of Pennsylvania study found that high school students using ChatGPT scored worse on tests, while a specialized AI tutor improved problem-solving but not test scores, highlighting potential learning inhibition.