AI Tools for Assignments and Research

Generative "Artificial Intelligence" isn't as "intelligent" as the name implies.

GenAI tools, or Large Language Models, are data sets paired with a series of pre-programmed behaviours, and they have several known limitations and problems, including:

Hallucinating

Some tools, like ChatGPT will provide a "likely" responses based on their data sets, even though those responses are not correct. They have been known to "generate" fake references to go with the fake information they have also generated.

Test the tool: Before asking about something you aren't familiar with, try asking it to write about something you know well. Ask it to provide references, and see if you can find all of the references it provides.

Gaps in details

For programs that are fed with large amounts of information, LLMs often omit details for no apparent reason.

Test the tool: Before asking it to summarise or explain a process you are unfamiliar with, ask it to outline a process you can check easily, like the rules of a game or the plot of a book or movie.

Gaps in perspectives

GenAI tools/LLMs are largely trained on material from a Western perspective – predominantly upper-middle class, white, male and American. Perspectives from minority groups and other countries are largely missing from the datasets used by the LLMs to generate their answers, so you might not see the most relevant perspective of an issue.

It is important to make sure you do not assume the content is culturally neutral because it was generated by a machine.

Bias and ethics

Always read the responses given by a generative AI tool critically.

You should also be aware that the "ethics" of a computer program are based partly on the instructions and parameters set by the programmers, partly on the content of the data used to train the program, and partly on machine logic. This can be result in unpredictable and ethically "dubious" ethical standards being applied to the content the program generates.

You are responsible for the content you submit.

If you are permitted to use GenAI for any part of your assignment, you must still check it for accuracy and bias. You are responsible for the content you submit, and while you may be permitted to "outsource" some aspects of the assignment preparation to the tools, you are still expected to vet the work in an intelligent and meaningful way before submitting it.