Assessment

A note on detection tools

Though Turnitin initially enabled its AI detection in Canvas assignments for Georgetown University, it has been disabled primarily because the tool has been reported to identify sentence-level false positives at a rate of 4% or higher, and has proven to falsely identify AI-written text at much higher rates in many other discrete instances (D’Agostino, 2023; Chechitelli, 2023). As Turnitin’s Chief Product Officer noted in a blog post, the detection tool has been designed to provide a “flag report,” serving as an “indicator” rather than a confirmation that language in an assignment may have been written by AI (Chechitelli, 2023). As LLMs become increasingly sophisticated, the accuracy and precision of detection software decreases (Sadasivan et al., 2023).

Additionally, “instances of false accusations of AI usage being leveled against students at other universities have been widely reported over the past few months, including multiple instances that involved Turnitin (Fowler, 2023; Klee, 2023).” One of the harmful side effects of the detection software’s inaccuracy has been impacting non-native English speakers (Sample, 2023; Myers 2023; Liang et al, 2023). Other universities—see Vanderbilt University and the University of Pittsburgh—have also disabled Turnitin’s AI detector. The Center for New Designs in Learning and Scholarship advises against relying on an AI detector to assess student work, and set forth clear expectations about how AI tools can or cannot be used in assessments/assignments.

Bias and inclusivity

We all bring biases with us into the classroom (Banaji & Greenwald 2013); and so do AI tools, maybe even more so. See the resources below which expand on the ways AI tools may amplify biases, and how to address and mitigate these biases to whatever extent possible:

Data and privacy

Among the many complications surrounding data and privacy in generative and detection (see below) AI tools, one issue prominently arises: “when plagiarism detection services archive student work… student privacy rights may be violated,” (Brinkman 2013). Do not upload any student work to AI tools (generators or detectors) without their permission. It’s important to “consider whether the information about students shared with or stored in an AI-enabled system is subject to federal or state privacy laws, such as FERPA,” (Department of Education 2023).

As you talk with your students about data and privacy implications, you may want to review the privacy policies you (and students) are agreeing to by engaging with the tools. For example, in teaching with Chat GPT, read over the porous privacy policy with students and discourage them from sharing personal information on the platform. Note that:

  • The company may access any information fed into or created by its technology.
  • They use log-in data, tracking, and other analytics.
  • Their technology does not respond to “Do Not Track.”

Allow students to opt out if they don’t feel comfortable having their data collected. See Open AI’s full policy here and these FAQs on how the company may use information shared with it.

References

Banaji, M.R. & Greenwald, A.G. (2013). Blindspot: Hidden biases of good people. New York: Delacorte Press.

Brinkman B. (2013). An analysis of student privacy rights in the use of plagiarism detection systems. Science and engineering ethics, 19(3), 1255–1266. https://doi.org/10.1007/s11948-012-9370-y.

Chechitelli, A. (2023). AI writing detection update from Turnitin's Chief Product Officer. Turnitin. turnitin.com/blog/ai-writing-detection-update-from-turnitins-chief-product-officer.

Coley, M. (August 2023). Guidance on AI Detection and Why We’re Disabling Turnitin’s AI Detector. Vanderbilt University Center for Teaching. https://www.vanderbilt.edu/brightspace/2023/08/16/guidance-on-ai-detection-and-why-were-disabling-turnitins-ai-detector/.

D’Agostino, S. (June 2023). Turnitin’s AI Detector: Higher-Than-Expected False Positives. Retrieved June 1, 2023, from https://www.insidehighered.com/news/quick-takes/2023/06/01/turnitins-ai-detector-higher-expected-false-positives.

Dartmouth College Writing and Rhetoric Program. (n.d.). Syllabus and Assignment Design. Retrieved from https://writing.dartmouth.edu/teaching/first-year-writing-pedagogies-methods-design/syllabus-and-assignment-design.

Fowler, G. A. (April 2023). We tested a new ChatGPT-detector for teachers. It flagged an innocent student. Retrieved April 4, 2023, from https://www.washingtonpost.com/technology/2023/04/01/chatgpt-cheating-detection-turnitin.

Frances, C. and Zimotti, G. (January 2023). Robots vs. Humans: Does ChatGPT Pose a Challenge to Second Language Writing?. The FLTMAG. https://fltmag.com/chatgpt-second-language-writing/.

Gally, Tom. (2023). Using ChatGPT for Language Learning [Video]. YouTube. https://www.youtube.com/watch?v=l41hZLRsDos.

Georgetown University Writing Program. (n.d.). Writing in the Majors. Retrieved from https://writing.georgetown.edu/writing-in-the-majors/.

IBM. (2023). What is natural language processing?. https://www.ibm.com/topics/natural-language-processing.

Intriligator, James. (July 2023). "How to Get the Best Results from ChatGPT." Tufts Now. https://now.tufts.edu/2023/07/21/how-get-best-results-chatgpt.

Julia Staffel. (September 2023). Chat GPT and its Impact on Teaching Philosophy and Other Subjects [Video]. YouTube. https://www.youtube.com/watch?v=bkjVkfU9Gro.

Klee, M. (June 2023). She Was Falsely Accused of Cheating With AI — And She Won’t Be the Last. Retrieved June 7, 2023, from https://www.rollingstone.com/culture/culture-features/student-accused-ai-cheating-turnitin-1234747351.

Korinek, Anton. (February 2023). Language Models and Cognitive Automation for Economic Research. National Bureau of Economic Research. https://www.nber.org/papers/w30957.

Kortemeyer, G. (2023). Can an AI-tool grade assignments in an introductory physics course?. arXiv. https://arxiv.org/abs/2304.11221.

Li, Belle & Bonk, Curtis & Kou, Xiaojing. (2023). Exploring the Multilingual Applications of ChatGPT: Uncovering Language Learning Affordances in YouTuber Videos. International Journal of Computer-Assisted Language Learning and Teaching. 13. 1-22. https://dl.acm.org/doi/abs/10.4018/IJCALLT.326135.

Liang, W., Yuksekgonul, M., Mao, Y., Wu, E., & Zou, J. (2023). GPT detectors are biased against non-native English writers. https://arxiv.org/abs/2304.02819.

Masters, Ken. (2023) Ethical use of Artificial Intelligence in Health Professions Education: AMEE Guide No. 158, Medical Teacher, 45:6, 574-584, https://doi.org/10.1080/0142159X.2023.2186203.

MLA-CCCC Joint Task Force on Writing and AI Working Paper No. 1. (2023). Retrieved from https://hcommons.org/app/uploads/sites/1003160/2023/07/MLA-CCCC-Joint-Task-Force-on-Writing-and-AI-Working-Paper-1.pdf.

Myers, A. (May 2023). AI-Detectors Biased Against Non-Native English Writers. Stanford HAI. https://hai.stanford.edu/news/ai-detectors-biased-against-non-native-english-writers

Sadasivan, V. S., Kumar, A., Balasubramanian, S., Wang, W., & Feizi, S. (2023). Can AI-Generated Text be Reliably Detected? (arXiv:2303.11156). arXiv. https://doi.org/10.48550/arXiv.2303.11156

Saravia, E. (2022). Prompt Engineering Guide. Democratizing Artificial Intelligence Research, Education, and Technologies. https://github.com/dair-ai/Prompt-Engineering-Guide.

Sample, I. (July 2023). Programs to detect AI discriminate against non-native English speakers, shows study. Retrieved July 17, 2023, from https://www.theguardian.com/technology/2023/jul/10/programs-to-detect-ai-discriminate-against-non-native-english-speakers-shows-study.

The Associated Press. (2023, July 20). Standards Around Generative AI. Retrieved from https://blog.ap.org/standards-around-generative-ai.

Tiulkanov, Aleksander. (January 2023). Figure 1 in ChatGPT and Artificial Intelligence in Higher Education: Quick Start Guide. UNESCO. https://unesdoc.unesco.org/ark:/48223/pf0000385146.

U.S. Department of Education, Office of Educational Technology. (2023). Artificial Intelligence and Future of Teaching and Learning: Insights and Recommendations. https://www2.ed.gov/documents/ai-report/ai-report.pdf.

University of California, Berkeley. (n.d.). Understanding AI Writing Tools and Their Uses in Teaching and Learning. Retrieved from https://teaching.berkeley.edu/understanding-ai-writing-tools-and-their-uses-teaching-and-learning-uc-berkeley.

University of Pittsburgh. (June 2023). Teaching Center doesn’t endorse any generative AI detection tools. University Times. https://www.utimes.pitt.edu/news/teaching-center-doesn-t.

University of Maine. (2023). Create an interactive game to practice a language [Video]. https://nmdprojects.net/learnwithai_www/media/gpt4_practice_foreign_language.mp4.

Vee, A., Laquintano, T., & Schnitzler, C. (Eds.) (2023). TextGenEd: Teaching with Text Generation Technologies. The WAC Clearinghouse. https://wac.colostate.edu/repository/collections/textgened/.

Wei, Y. (2023). Repilot: iSE-UIUC. https://github.com/ise-uiuc/Repilot (Original work published 2022).