We have all read about cases in which lawyers apparently relied on generative artificial intelligence systems to produce arguments and citations that can be (and have been) submitted to courts. When a judge gets a filing that seems to have citations that don’t exist, they take it seriously. But there is a danger if we think that false statements made by generative AI systems are limited to these systems being directed to provide content for legal briefs.
The growing use of generative AI in all contexts in business and governmental organizations means that the dangers associated with what has come to be known as “AI hallucination” (that is, instances in which AI systems generate content that is held out to be true when, in fact, it is not.)
For your paper, you will be in the position of a CIO who has come to recognize the risks associated with AI hallucinations. Assume you are writing a document that the General Counsel would send out to all of the senior managers of your company/agency.
- Begin with an introduction in which you provide your definition of AI hallucination and why you believe it is a real problem worth addressing. This should not exceed 1.5 pages.
- Then provide three examples of how applications of generative AI might be used in an organization that you believe would cause difficulties if they generated content with AI hallucinations that were not caught before the text was used. For each, please briefly explain the application, what falsehoods might be generated, and what steps you believe would be appropriate to defend against false information generated by generative AI systems from being released. Each of your three examples should be between 1 and 1.5 pages.
- Finally, provide your conclusion as to how the organization should approach the problem of AI hallucination, and steps that you believe would be reasonable, doable and which would be realistic in light of resource requirements for your solution. This should not exceed 1.5 pages.
Your paper should be clearly written. Your completed paper should not exceed five 1.5-spaced pages. Think beyond the legal effect of signing and consider how we could make sure that people understand what they sign, whether you favor the concept of informed consent or not.
Please use properly formatted citations (APA) to reference information from our assigned readings, or from your own online research. Note, however, that citations from the assigned readings won’t count as research. Our expectation is that you will use the assigned readings as a starting point but will carry out independent research beyond the assigned readings. That research is documented through citations in the text, not just a readings list.
As with all the assignments for this course, there is an expectation that you can express yourself effectively in writing. Put another way, spelling, and grammar count. Feel free to use spell-checkers and/or grammar-checking tools. However, you may NOT use any form of generative artificial intelligence (AI) to prepare your paper, even if you plan to use it to generate a draft and then edit it. Also, remember that generative AI has been found in the past to create nonexistent references. You definitely don’t want to cite references that aren’t real and reviewable.