Many of my colleagues fear ChatGPT. They fear that students will use this new online tool based on artificial intelligence to generate papers and answer questions on homework assignments that students are meant to tackle themselves. Their fear is legitimate. Why? Because if you read ChatGPT’s answers to a range of questions, you notice something rather quickly. ChatGPT generally produces plausible, thoughtful, coherent responses. Furthermore, ChatGPT writes well — at least compared to what I see from most of my undergraduate students. In short, ChatGPT creates headaches for instructors. It is forcing us to reconsider how we evaluate our students.
On the other hand, I was excited to learn what this new AI tool would produce when asked about loon biology. I asked ChatGPT: “Do loons mate for life?”. It responded:
While common loons often return to the same breeding territories and may reunite with the same mate from the previous year, these pair bonds are not necessarily permanent. After the breeding season, loons typically migrate to coastal or open water areas for the winter, and their social behavior can be more flexible during this time. When they return to their breeding territories in the following spring, they may or may not reunite with the same mate.
Now, this is not a terrible answer. ChatGPT is correct when it suggests that loons do not mate for life. But its mention of migration is off-topic. Its point regarding “social behavior” during winter is irrelevant. Its statement that loons “may or may not reunite with the same mate [in the spring]” is grossly misleading. And ChatGPT has entirely failed to mention the chief reason why loons do not mate for life: territorial eviction.
ChatGPT’s responses to loon-related questions are similar to its answers on other topics. The program searches the web for information from which to build its responses. But because these responses draw information from countless online sources, they include the good, the bad, and the ugly. In short, ChatGPT gives you a blur of “conventional wisdom” about any topic, not rigorous and precise scientific truth. Its answers are, for the most part, uncontroversial, vague, and riddled with caveats.
I turned my attention to loon conservation. When I asked, “Are any common loon populations in trouble?”, ChatGPT came back with:
As of my last knowledge update in September 2021, the status of loon populations, particularly the Common Loon (Gavia immer) in North America, varied by region and was a subject of concern in some areas. Common Loons were classified as a species of “Least Concern” by the International Union for Conservation of Nature (IUCN) due to their relatively stable population at the global level.
However, it’s important to note that local and regional populations can face significant challenges, including habitat loss, water pollution, lead poisoning from ingesting fishing tackle, disturbance from boating and recreational activities, and climate change impacts on their breeding and wintering grounds. These factors could affect specific loon populations and lead to localized declines.
ChatGPT is spot on in some respects. Loons are in no trouble globally. And the status of populations varies from one region to another. But the program uncritically lists many anthropogenic factors that might harm loon populations and provides no specific information. One might infer from ChatGPT’s statement that we must solve all of the problems listed in order to preserve loons. Or perhaps loon populations face many obstacles but none is severe enough to cause serious harm. In truth, we do not know which, if any, of the listed factors might endanger loons. Furthermore, ChatGPT has failed to mention two factors — exploding black fly populations and decreased water clarity — known to harm loon populations in the Upper Midwest. If this had been a class assignment, ChatGPT would have earned a “C-“.
ChatGPT’s take on factors threatening loons sounds eerily familiar to what we loon conservationists often say. (I include myself in this group.) That is, we often settle for a ChatGPT-caliber pronouncement about loon conservation. We say or imply that if we: 1) enact a lead ban, 2) reduce human settlement on northern lakes, 3) eliminate water pollution, 4) keep boaters away from adults and chicks, and 5) halt climate change, then we can save loons. These goals, of course, range from difficult to impossible to achieve.
A ChatGPT-like approach will not save loons. We simply cannot eliminate all threats. In fact, many factors that loon conservationists perceive as threats pose little or no hazard. The best example is human recreation. While lead sinkers are a danger to loon populations because of the many breeding adults they kill each year, loons are well-equipped to handle boaters. Loons tremolo, wail, or yodel when boats are too close to them or their chicks. They penguin dance, charge, and dive noisily at boats that creep close to their nests. Humans usually interpret these warnings correctly and back away in short order. Moreover, the low rates of mortality and nest abandonment attributable to boating activity indicate that loons tolerate human interactions well.**
How do we distinguish between relatively benign environmental factors, like boating, and those that truly endanger loon populations, like black flies, lead sinkers, and water clarity? Science. Only science will save loons. If the Minnesota loon population is declining and 50% of loon deaths result from lead poisoning, the folks in Saint Paul will take notice. If Wisconsin loons fledge 40% fewer offspring nowadays than in 1995 because of lower water clarity, that will make headlines in Milwaukee. And if the fledging rate across the Upper Midwest is 30% lower owing to black fly-induced nest abandonments, that might turn apathy concerning climate change into action.***
So if you are an unprincipled student being taught by an inattentive instructor, ChatGPT might earn you a “B” on your history paper. And if you cannot get the wording right in a letter to a client or colleague, ChatGPT might provide suggestions. But we loon conservationists should resist the uncritical, shotgun approach that ChatGPT takes to addressing questions. If we are going to preserve loon populations that we treasure, we must first collect data. And then, in our communications with the public, politicians, and the media, we must highlight the specific environmental hazards that science has shown to be dangerous to loons.
* The featured photo was taken a week ago by Linda Grenzer of an adult male from Lake Winter, Wisconsin that swallowed a lead sinker. Though the sinker was surgically removed by Raptor Education Group in Antigo, the male died two days later from the lead it had already absorbed.
** I might be biased by my work in Wisconsin and Minnesota in my conclusion about loons coping with boating activity. It is conceivable that boats do cause enough nesting failure and/or chick mortality in some regions to threaten loon populations. We do not have data to show such a pattern, however. If you know of such data, please let me know.
*** I made up these three numbers. We are still collecting data on loon mortality caused by lead sinkers. Only recently have we learned about the threats of black flies and loss of water clarity. We and our partners are recording many lead-poisoned loons these days. We will publish an estimate of lead’s impact within five years. Robust estimates of the impacts of black flies and water clarity on loon populations in Wisconsin (and possibly Minnesota) should be available by sometime in 2025.