What are the ethical ramifications of using GenAI?
Answer
There are several ethical concerns about GenAI. Sweetman (2023) has created a visualization of “Some Harm Considerations of Large Language Models (LLMs).” These include biased data, resource intensiveness (carbon footprint and water usage), copyright and privacy concerns, exploitative labour, and inequity or access barriers.
There are also ramifications that can affect the integrity of you and your assignment, even if you use GenAI with the best intentions. Its output can be biased according to the materials it was trained on, thereby creating potential bias in your assignment. Seek out multiple perspectives, and use your critical thinking to determine whether a certain viewpoint is being promoted. Consider what your own viewpoint is, and what scholarly evidence supports it. Overall, seek to use AI in ways that pose the least risk of reproducing stereotypes or specific ideologies; this means it should be an assistant or tool, not a researcher or writer.
Furthermore, much of the information GenAI is trained on is derived from authors without their consent. Therefore, you cannot properly cite where the information originally came from. This is a reason to avoid treating GenAI as a reliable source. Rather than quoting or paraphrasing from it, treat it like an advanced search engine that leads you to other sources. Then evaluate those sources, or seek scholarly sources that you can trust and properly acknowledge.
- Learn about evaluating web sources for reliability to ensure the information is accurate.
Another concern is that GenAI can fabricate references, meaning it creates fake articles with authentic-sounding authors and titles. Do not rely on LLMs such as ChatGPT to do your research for you; use it only as a starting place for background reading or preliminary searching. Always review the original source before citing it. You can also get research support by contacting the library.
Finally, be sure to turn off chat history and model training to protect your own privacy when using an LLM. Otherwise, it may use the data you input to train its model.
- Learn about acceptable uses of AI and some alternatives.
Reference
Sweetman, R. (2023). Some harm considerations of Large Language Models (LLMs). eCampusOntario H5P Studio. https://h5pstudio.ecampusontario.ca/content/51741
French version: https://etsmtl.libguides.com/IAgenerative/biais