AI evidence extraction

ChatGPT in education: Strategies for responsible implementation

Mohanad Halaweh · Contemporary Educational Technology · 2023

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

5/10

Relevance

2/4

Quality (LMQS)

Evidence

576

Citations

20.47

FWCI

Top 10%

Impact

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.30935/cedtech/13036

Methodology & findings

Study design

Narrative literature review with argumentative analysis.

Main result

The paper argues that "ChatGPT produces outputs of a high quality that have a high probability of passing plagiarism detection software" and that "AI contents detector tools exist, which can detect with high degree of accuracy whether text has been generated by a human or an AI such as one used by OpenAI." Specifically, "Turnitin (2023) announced that it developed an AI tool that identifies 97% of ChatGPT and GPT-3 authored writing." The authors conclude that "universities should take a proactive rather than a reactive approach, and adopt AI technology in the realm of education, learning, and assessment" rather than banning or blocking ChatGPT.

Research paradigm

interpretivist/argumentative

Author conclusions

The authors conclude that "By combining ChatGPT and human authors, the output is superior in terms of creativity, originality, and efficiency than if either one was to work alone" and that "Allowing students to use the tool gives them an equal chance to develop ideas and improve their writing, as it is encouraged by the faculty." Furthermore, they state "Universities and instructors are encouraged to consider the suggested policy, modifying or extending it to fit the individual needs of their institutions and courses." Finally, they note that "Presentation/viva and defending one's work will become standard assessments in the educational environment, in order to verify the learning specially when assessment are done in collaboration with ChatGPT."

Risk of bias

Author advocacy bias: The paper explicitly argues in favor of ChatGPT adoption, potentially biasing discussion toward supportive interpretations; Selection bias: Literature review is not systematic; examples chosen appear selective to support the pro-adoption argument; Publication bias: Limited peer-reviewed literature available at time of writing (3 months post-launch); Temporal bias: Knowledge cutoff of ChatGPT training data (2021) and rapid technology evolution limit validity; No control group or comparison conditions; Author bias: Single-authored argumentative piece presenting one perspective on ChatGPT adoption; Selection bias: Limited peer-reviewed literature available at time of writing (3 months post-launch); Confirmation bias: Paper develops argument in favor of ChatGPT use, potentially selecting supporting evidence; Temporal bias: Knowledge base of referenced ChatGPT reflects training data only through 2021; Selection bias in literature review: limited peer-reviewed sources available at time of writing; Author position: strong advocacy for ChatGPT use may influence interpretation; Temporal limitation: knowledge cutoff of ChatGPT (data until 2021) noted but not systematically addressed; Scope limitation: focuses on specific educational concerns rather than comprehensive AI concerns

Open questions raised

The paper identifies that further development of ChatGPT by OpenAI could result in 'more sophisticated models and features as well as a higher level of accuracy due to the advancements in AI technology,' and proposes that 'OpenAI could create a specialized version tailored to academia (ChatGPT Academia).' The authors also note the need for training and education of faculty and students on proper use of the tool, and call for universities to integrate ChatGPT with learning management systems.
The paper identifies the following gaps and future directions: (1) Limited peer-reviewed research on ChatGPT in education at time of writing; (2) Need for faculty and student training on ChatGPT functions, accuracy evaluation, and query tracking; (3) Potential development of specialized ChatGPT Academia version tailored to academic contexts similar to Google Scholar; (4) Adoption of presentation/viva assessments as standard practice to verify learning in AI-augmented contexts; (5) Integration of ChatGPT with learning management systems; (6) Development of institutional policies adapted to individual institutional and course needs.
The authors identify that there is limited peer-reviewed research on ChatGPT in education at the time of writing. They suggest future development of "a specialized version tailored to academia (ChatGPT Academia)" and propose that the techniques outlined in the paper "may be adopted for the purpose of academic research and publication by journals." The paper also suggests training needs for faculty and students on proper ChatGPT use.

Extracted from: pdfAgreement 73%

Explore related topics

Responsible AI Use Doctoral Training Position Paper