Classical rhetoric arose as a discipline designed to aid speakers before courts of law and other public settings. By identifying three primary means of persuasion—appeals to authority (ethos), appeals to emotion (pathos), and appeals to reason (logos)—it taught speakers to calculate more precisely which combination of these three would produce the desired effect in an audience. In its more modern form, especially after the scientific revolution, it has become one tool for teaching the delicate art of weighing evidence, whether present in spoken or written form. In a few short months, the new world of AI is eroding some of the foundations for this system, leading to an inevitable crisis of trust.
A large part of trust rests in the authority and reliability of evidence. This process has always been an art, not a science, because it operates as a form of prudential wisdom based upon principles, not universal maxims. The basic distinctions between primary, secondary, and tertiary sources are key here, and it is helpful to see how they developed from courts of law, which helps explain why they matter.
For example, if a person is shot, an eyewitness would be an ideal form of evidence. Is an eyewitness always reliable? No, but if an eyewitness is available, that evidence is incredibly valuable. The eyewitness testimony of classical rhetoric is equivalent to what today’s rhetoricians and researchers call “primary sources.” Primary sources are the gold standard for reconstructing a past historical event. Those primary sources could be written accounts, photos, videos, or other kinds of physical proof, like archaeology. Classical rhetoric distinguishes eyewitness testimony from physical evidence, but there is something to the idea that the value of both lies in the proximity of the evidence to the event. Primary sources carry a great deal of ethos or authority when it comes to figuring out what happened in a historical event.
One of the most disconcerting aspects of AI concerns the way it undercuts the reliability of primary sources. While it has always been possible to doctor photographic and videographic evidence—the Soviet Union did it all the time—AI now makes the modification and complete fabrication of primary source evidence easier by the day. Increasingly, it becomes harder to trust any source.
My father remembers a time when people began to doubt the historical fact of the Holocaust. When he visited his small-town barber as a young teenager, the barber—who had been in Europe during WWII—pulled out a shoebox of photos he had taken when they liberated the concentration camps. Note that the presence of the photos combined with the testimony of a trusted eyewitness strengthened my father’s belief that the event had indeed occurred. Increasingly, with the rise of AI, knowing the trustworthiness of a source for photo and videographic evidence will become key.
Another kind of testimony that might be brought before a court of law in the case of a shooting would be an expert witness. This might be someone who understands the nature of things involved in the case. Perhaps the person is an expert in guns or psychology or security. An expert witness may not have been at the event, but their expertise allows them to put together the facts of the case in a particular way. Perhaps there is something about the eyewitness testimony that does not correspond to the science of shot trajectories. Perhaps the pieces of evidence can be put together in a way that leads to an unforeseen conclusion.
This kind of expert witness is analogous to what is called a secondary source. A secondary source is typically an analysis of the primary and secondary source evidence by an expert in their field. Academics call this kind of writing scholarship, including articles, book chapters, and academic books. Notice how ethos here derives from the background of the expert and the logos (logical argumentation) of that person’s scholarship. Students often assume that secondary scholarship is by definition authoritative, but a large part of academic training consists of helping students to analyze the logical argumentation of the expert themselves.
The sheer capabilities of AI can make it appear to be the equivalent of an expert witness or secondary source. On the other hand, experts in various fields note that this technology can make basic mistakes because it lacks information or cannot distinguish between the relative reliability of sources. This inconsistency of product can make AI’s results less trustworthy than Wikipedia. As trust in academia wanes and AI gives the appearance of being an “expert,” evaluators of evidence increasingly have to make hard decisions about the reliability of the sources. While it can be tempting to treat AI like a secondary source, it lacks the prudential wisdom found in human beings necessary to consistently evaluate the relative trustworthiness of its own sources. Could it be a tertiary source instead?
Tertiary sources can vary quite a bit in their reliability and authority. In a court case, the closest equivalent might be the summary of the arguments given. The quality and reliability of summaries can vary quite widely depending upon who is summarizing, how much they understood of what took place, and how well they are able to summarize. A trained judge might summarize the results of a court case quite differently from a casual observer.
In academic scholarship, encyclopedias are a good example of tertiary sources. The reliability and accuracy of encyclopedias can vary quite widely, even within the same encyclopedia. An edited encyclopedia for a general audience, like the Encyclopedia Britannica or World Book, will likely give an account of a subject from the perspective of “common knowledge.” This is one reason that these works are generally cited in elementary school papers but not college-level papers. In elementary school, a teacher wants to know where a student learned their information, but a college student can be expected to write about topics of “common knowledge” without needing to cite. An academic encyclopedia is often subject-specific and written by experts on that subject. One major feature of an academic encyclopedia is the use of primary and secondary source references that enable a reader to research further.
Wikipedia occupies an interesting space when it comes to tertiary sources because sometimes it gives information that is common knowledge, but it also encourages authors to give citations. The tricky aspect of Wikipedia is the lack of editorship, meaning that articles can vary quite a bit in quality. One may be written by an expert in Beowulf who writes just for the fun of it while the other could be written by someone who just thinks they know something about WWII. For this reason, most people use it but never cite it. The best users of Wikipedia and any other tertiary source look up the citations given in order to evaluate and substantiate the claims in the article. The inconsistency of the medium means that it cannot be trusted as a source, merely as a pointer to other sources.
Perhaps one of the best antidotes to this particular problem is to train students to treat AI like a tertiary source by:
- always looking up the links in an AI summary and
- not limiting their reading to merely the articles mentioned in the summary.
This kind of education and training of students is made difficult, however, by the visual presentation of AI summaries because ethos can come from both substance and appearance. Just as a doctor in a white lab coat looks authoritative, even if they don’t actually have a degree, an AI summary physically presents itself as the answer. The physical presentation actually undermines the attempt to educate students because AI looks so, well, authoritative.
When viewed from the perspective of primary, secondary, and tertiary sources, AI increasingly presents scholars with what can only be termed an ethos crisis. It took centuries to develop systems of trust based upon evaluating sources that AI is destroying in only a few months. While the discipline of rhetoric cannot solve all the problems that AI presents to the modern student and teacher, its categories remain clarifying, helping us to see a few reasons why we have a crisis of authority today, a crisis that worsens as AI improves.
