Word on the street is that artificial intelligence is here to stay. Some of us are wont, with these breakthroughs, to quote that unusually profound Hollywood gem, “Your scientists were so preoccupied with whether or not they could, they didn’t stop to think if they should.” But philosophy was hardly ever a match for capitalism—and so we grudgingly adapt.

We have yet to see all that artificial intelligence will unleash. In the short term, at least, we can form a sense of its effects on knowledge transmission and academic research. Through our own prompts, we can use it to test what it offers to students, teachers, casual researchers, and professionals who seek quick access to a specific topic. AI aggregating platforms like ChatGPT synopsize available information. Experts can verify what various groups are gleaning from their field and whether there are deficiencies. Those deficiencies might be in the research itself, but also in failures to disseminate the most up-to-date and accurate information.

In the case of Franco-American history, AI platforms show some promise as well as clear limitations, as I found when putting ChatGPT to the test. To begin with, I asked ChatGPT, “Write a short history of Franco-Americans.” The 300-word answer was, all told, a reasonable overview. I reproduce it here in its entirety:

The Franco-American community has a rich and diverse history rooted in the migration of French-speaking people to the United States.

Early Migration:

The story begins in the 17th and 18th centuries when French explorers and settlers, including those from Canada (New France), ventured into what is now the United States. These early French settlers established communities in areas like Louisiana, where French culture had a significant influence on the region, and in parts of the Midwest and the Northeast.

19th Century Migration:

The most substantial wave of Franco-American migration occurred in the 19th century. In the 1800s, many French Canadians, seeking better economic opportunities and escaping political unrest in Quebec, moved south to New England and other parts of the U.S. They settled in industrial cities like Manchester, New Hampshire, and Lowell, Massachusetts, where they found work in textile mills and other industries. These immigrants maintained strong ties to their French-Canadian heritage, creating vibrant communities with their own churches, schools, and cultural organizations.

Cultural Contributions:

Franco-Americans have made significant contributions to American culture and society. In Louisiana, the Cajun and Creole cultures have roots in French colonial history and have greatly influenced the state’s unique cuisine, music, and festivals. In New England, Franco-American communities have preserved traditions through festivals, music, and cuisine, contributing to the rich tapestry of regional culture.

20th Century to Present:

Throughout the 20th century, Franco-Americans have navigated a complex identity, balancing their French heritage with their American experiences. Many have assimilated into broader American culture, but there has been a resurgence in recent decades of interest in preserving and celebrating Franco-American heritage. This has included efforts to revive the French language, celebrate cultural festivals, and maintain historical connections.

Today, Franco-Americans continue to be an integral part of the American cultural landscape, with a legacy that highlights the diverse and dynamic nature of the American experience.

We can nitpick, but it would be difficult to create a significantly better overview that shows the same geographical and temporal ranges within the same word count. In terms of sheer coherence, the answer outshines some Wikipedia articles on Franco-Americans. Beyond this, the (current) promise of AI begins to unravel.

I next asked ChatGPT to provide sources for the above answer. The platform returned two books and two academic articles—none of which I recognized, and I have been conducting in-depth research on Franco-Americans for twelve years in addition to publishing a large bibliography on the subject. I did not even recognize the authors. To the best of my knowledge, all of these works, including the first, Marc G. R. Broussard’s The Franco-Americans of New England, are invented. Also listed were three online resources. One was the Library of Congress’s American Folklife Center. The other two were a Cajun author’s personal website and the site of the New England Historical Society, which is not a historical society and, despite the pretense of authority, does not offer the rigor or resources expected of a historical society. The final two sources were the French Canadian Historical Society of Illinois (which, to the best of my knowledge, does not exist) and the Franco-American Centre at the University of New Hampshire (which definitely does not exist). Wikipedia has ChatGPT beat when it comes to easy access to sources, it would seem.

I tried the question differently: “What are the most influential books, chapters, and articles on the history of Franco-Americans?” Many of the same fictitious works and dubious resources were repeated. ChatGPT appeared to mimic existing sources, aiming for verisimilitude. One French name (Marc Broussard) stood in for another (Yves Roby); John P. DuLong became the non-existent John H. R. Long. ChatGPT named actual academic journals—but journals that had never carried the articles in the answer.

Knowing that the program needs hand-holding, I went further: “Many of the sources you provided above do not actually exist. Provide real, published sources that can be found in Worldcat, in academic journal databases, and on reputable websites.” The result was the same. I also asked for a synopsis of Yves Roby’s Franco-Americans of New England: Dreams and Realities. The answer was vague and bland; it might well have been a synopsis of Armand Chartier’s work, or Gerard Brault’s, or just about anyone else’s.

It became clear that, insofar as Franco-American history is concerned, ChatGPT works best when it comes to the big picture. When asked to name major events in the history of Franco-Americans, it gave the following with very brief descriptions:

• Exploration and Early Settlements (1600s-1700s)
• Great Migration from Quebec (Late 1800s-1910s)
• Formation of Franco-American Communities (Late 1800s-Early 1900s)
• The Rise of French-Canadian Organizations (Early 1900s)
• World War I and II (1914-1918, 1939-1945)
• Decline of Textile Industry (Mid-20th Century)
• Cultural Revival and Preservation Efforts (Late 20th Century-Present)
• Recognition of Franco-American Heritage (Late 20th Century-Present)
• Franco-American Heritage Month (2005-Present)

Few specifics were given for each of these “events,” with the exception of the last. “In 2005,” ChatGPT stated, “the U.S. Congress recognized September as Franco-American Heritage Month, acknowledging the significant contributions of Franco-Americans to the cultural and historical fabric of the United States.” This is inaccurate. French-American Heritage Month comes in July.

The program defines concepts (survivance, for instance) well enough. When asked to name prominent Franco-Americans, living or dead, it did successfully name, with appropriate descriptions, people of French heritage who have lived in the United States. The list included Louis Riel and Mike Gravel, but otherwise hardly anyone with French-Canadian roots or ties to New England and New York. Ferdinand Gagnon, Nap Lajoie, Eva Tanguay, Corinne Rocheleau-Rouleau, and Jack Kerouac all went unmentioned.

Again hoping for specifics, I asked for a short biography of Ferdinand Gagnon. The first paragraph of the ChatGPT answer tells the whole story of this experiment:

Ferdinand Gagnon (1885-1960) was a notable Franco-American figure known for his significant contributions to the preservation and promotion of Franco-American culture and history. He was a prominent leader in the Franco-American community and played a crucial role in advocating for the recognition and preservation of Franco-American heritage.

The description is not inaccurate, though it is extremely vague and Gagnon would not have spoken of Franco-American heritage per se. A proper encyclopedia entry would begin by mentioning specific achievements. But far more worrisome is the lifespan given. Gagnon died in 1886, not 1960. AI could not offer what a basic Google search reveals instantly. Conceptually, the overall answer is fair. Gagnon was a writer; he did play a leadership role within the budding Franco-American community. Yet: no mention of journalism or of Le Travailleur, which he operated and edited for more than a decade and which made his reputation.

Undeterred, I gave ChatGPT greater guidance: “According to the Dictionary of Canadian Biography, who was Ferdinand Gagnon?” The program was equally undeterred. It doubled down on the lifespan it initially gave, but now added Saint-Philippe as Gagnon’s birthplace. The DCB has him born in Saint-Hyacinthe—accurately.

The experiment continued with Eva Tanguay. Remarkably, ChatGPT could provide specific information about Tanguay’s career as a vaudeville entertainer. Her vital years (1878-1947) were given accurately, but not her places of birth and death. Her actual date of birth was off by a full month. In short, ChatGPT provided fictions for some of the most easily verifiable data points.

A few more examples. I asked the program to provide details of the Flint Affair, a battle over the principle of ethnic parishes that raged in Fall River in the 1880s. ChatGPT misinterpreted that as labor activism in Flint, Michigan—fair enough. But when asked to define the “Chinese of the East” controversy, a landmark battle of the early 1880s, the answer was somehow more disappointing. It read, “The term was coined by William A. MacDonald in his 1914 book, The French-Canadians of New England, and was used to highlight certain social and cultural aspects of the Franco-American community.” Sometimes, all you can do is either laugh or cry.

Some readers may interject with the Do’s and Don’t’s of AI. Prompt writers should provide context, define the audience, give the program a persona, etc. Abandoning short, direct prompts, I asked, “You must write a short essay on changes in the New England Franco-American community between 1945 and 1975. This will serve as context for a primary document that will be assigned to upper-level undergraduates. Provide clear examples and a short list of sources. The sources should be real and verifiable.” Well, I got more of the same.

I share all of this not to argue for the inherent futility of artificial intelligence—though the historian in me appreciates that the need to inculcate critical thinking and strong research skills is as present as ever. There may yet be a place for the human element. Nor is any of the above set in stone. AI platforms will continue to “learn” and become more exact. A researcher might enter the same prompts in several months and get considerably better answers.

For the time being, however, some things are quite clear. When it comes to Franco-American history, ChatGPT works well enough for brief assignments that are conceptual and relatively surface-level. When it comes to specifics, the program’s answers are inconsistent. If I were still working in a classroom setting, it would receive grades ranging from B+ to F. To an uninitiated user, it is not instantly apparent that if the program doesn’t “know,” it will invent. It is not a substitute for proper research and, although marketed as a way of disseminating knowledge more efficiently, only researchers with a strong foundation in the subject are likely to pick up on errors. Wikipedia, problematic though it is in many respects, at least offers sources—almost always real ones—to help guide research efforts.

Questions remain. There are abundant online sources about Gagnon and Tanguay. The fault is therefore in the program itself and not in its stars. But there is still much more we can all do to deepen research on Franco-American history and help disseminate factual information, such that, as artificial intelligence improves, eager learners can work from a reliable footing.