Given some of the major drawbacks of GenAI, should we refuse to use it?

Life isn’t that simple. Much of the technology we rely upon has GenAI embedded. But using GenAI should entail being aware, consciously, of its strengths and weaknesses and balancing its value versus costs in the context of the activities for which you intend to use it.

Models for Ethical Use

One model for such conscious or responsible use comes from health services, where efforts have been underway for several years now to ensure that all AI use is "trustworthy." By this, the US Department of Health and Human Services (2021) "refers to the design, development, acquisition, and use of AI in a manner that fosters public trust and confidence while protecting privacy, civil rights, civil liberties, and American values, consistent with applicable laws." Dimensions of trustworthiness–fairness/impartiality, transparency, robustness/reliability, accountability, privacy, and security–as well as simple examples of some of the means for securing each are shown below.

The image presents an "Overview of TAI Principles" that outlines six key principles for developing and using trustworthy artificial intelligence (AI) in government applications. The principles are arranged in a circular diagram with short descriptions:
Fair / Impartial: AI applications should include checks from internal and external stakeholders to ensure equitable application across all participants.
Transparent / Explainable: All relevant individuals should understand how their data is being used and how AI systems make decisions; algorithms, attributes, and correlations should be open to inspection.
Robust / Reliable: AI systems should have the ability to learn from humans and other systems and produce accurate and reliable outputs consistent with the original design.
Privacy: Individual, group, or entity privacy should be respected, and their data should not be used beyond its intended and stated use; data used has been approved by the data owner or steward.
Responsible / Accountable: Policies should outline governance and who is held responsible for all aspects of the AI solution (e.g., initiation, development, outputs, decommissioning).
Safe / Secure: AI systems should be protected from risks (including cyber) that may directly or indirectly cause physical and/or digital harm to any individual, group, or entity.
The text at the bottom notes that "TAI principles are not mutually exclusive, and tradeoffs often exist when applying them."
Alt text created by claude.ai — Source: HHS.gov

In higher education, we can build on these insights, remembering that accountability lies with the person who puts their name on the work.

Think of it this way: if you are planning a trip to Los Angeles, you might ask friends who’ve been there where they stayed and what they did. This would save you time in your preparations. But you would probably look into their recommendations to ensure they met your needs, were still in business, or open the day you might be there. Similarly, if you asked a friend how to get the cheese to stop sliding off of your pizza, you would not just accept it at face value if they told you to use 1/8th cup of non-toxic glue. Our relationship to GenAI output must be similarly critical.

That last example is no joke, but the product of a GenAI tool that Google released to US users in late May 2024. The New York Times traced this emergent recommendation back to an 11-year-old Reddit recipe post meant as a joke; likewise, a satirical post from the Onion was implicated in a recommendation to ingest pebbles to aid digestion and provide vitamins and minerals. Numerous other problematic answers were reported.

Although Google reacted swiftly to control the damage, the incident proved the truism "garbage in, garbage out." GenAI’s raw or unassessed output reflects the data on which a system is trained. Further, while some examples of this tool’s errors were funny, the potential for damage is clear. People with little practice in "critical thinking" are particularly vulnerable.