Leveraging AI technology for Provenance Research in Modern Chinese Ink Painting Collections

I recently worked on a project involving a collection of Chinese ink paintings and calligraphy that had been assembled over several decades, with some acquired in China before 1949, others obtained in Taiwan, and the rest collected in the United States from the 1970s to the 1990s. Almost all of the works were by well-known ink painters and calligraphers, several of whom are recognized a modern masters.

As is common with collections assembled in China and Taiwan during the mid-20th century, there was no written documentation of where they were acquired. The family lore was that the artwork was acquired directly from the artists. Typically, collections assembled in this period were obtained as gifts or tokens of appreciation from the artists. Given the prevalence of fakes, forgeries, and high-quality reproductions in the Chinese art market, proving a connection between the collector and the artist is crucial for verifying authenticity. The client, now in his 90s, wasn't sure if the three artworks by modern masters in the collection had been acquired through his family or his late wife's family. The country of origin for both families was China before 1949 and then Taiwan after that. He and his wife had met shortly after emigrating to the United States in the 1950s. Unusually, the surname of the family of the client was the same as his wife's family.

Provenance and Historical Context Research

Fortunately, the calligraphy provided a clue. Chinese paintings and calligraphy are often inscribed with dedications to the recipient from the artist. In this case, the calligraphy was dedicated to a Mr. Jingyuan. It was a slender thread on which to establish provenance, but it was a beginning. This is when I assumed the role of a detective and questioned the client about his father's and his wife's father's life dates, birthplace, career, education, and potential connections with prominent individuals of the time. In China during the early to mid-20th century, renowned artists frequently mingled with notable military and political figures, forming relationships that could be leveraged during those tumultuous and perilous times. Exchanging art was a means to solidify that relationship.

My client shared that his father was a physician and a military official in the Guomindang, the government led by Chiang Kai-shek in China, and later in Taiwan after 1949. His profession and military association fit well with the type of individual that would have been the recipient of art from well-known artists. The hitch was that the dedication to Mr. Jingyuan in the calligraphy did not align with the given name of my client's father. A dead-end? Perhaps, but perhaps not. In China, there was a cultural tradition among the educated elite, who were often amateur poets, calligraphers, and painters, to have multiple given names, including a courtesy name. Could it be that Mr. Jingyuan was an alternate name for my client's father?

An example of Chinese calligraphy

That is when I turned to AI to further my research. As professional researchers have all observed, AI has increasingly become a powerful tool for research, capable of scanning archives, exhibition catalogs, and auction records to collect relevant historical data. Additionally, AI can connect this information to specific artworks, offering a richer context.

I uploaded an image of the calligraphy. It correctly identified the artist through the signature and seals and summarized the calligraphy's content which was a poem. It also cited the dedication to Mr. Jingyuan. I entered a prompt containing the Chinese characters for the client's surname and the given name Jingyuan, and asked if the calligrapher was familiar with someone having that surname and given name. The response suggested that the name Jingyuan might be a variation of Gengyuan, a known associate of the calligrapher. However, this name did not match the client's father's name, Aiguo (altered here for privacy), and the explanation was unconvincing regardless. I then considered it from a different angle: was there any historical record linked to the client's father's name, Aiguo? There was not. I then inquired about a historical record for Jingyuan. The answer was affirmative. AI provided a promising summary of Jingyuan's biography, but I needed solid proof - access to the original source. AI provided a link to the Directory of Chinese Physicians affiliated with the Guomindang, published in Taiwan around 1950. The link led me to a record of a man with the same surname as my client and the given name Jingyuan, sharing the same life dates as my client's father. The document outlined his education, professional accomplishments as a medical doctor and professor of medicine, military rank, and notably, an alternate name: Aiguo. I was thrilled to establish a strong connection. However, I was slightly worried that it was a transcription instead of the original record, so I made initial attempts to access the historical archives but was unsuccessful. I then sent the transcription to my client, asking him to confirm if the professional titles and military rank matched what he knew about his father. They did. At this point, I felt confident that Mr. Jingyuan named in the dedication was the same person as my client's father.

Challenges and Limitations

While AI provided valuable research assistance guiding me to original sources that would have previously necessitated contacting archivists and experts or traveling to far-off libraries or archives, there were numerous informational wrong turns and missteps supplied by the AI search along the way. I needed to treat all the provided information with skepticism and modify my questions as necessary. During my research, I experimented with various platforms, including Gemini, Chat GPT, and Claude. I found that Gemini was better equipped for the type of inquiry I was conducting and was more adept at handling images of Chinese calligraphy and Chinese sources than the other two. Lastly, I had to go back to my client, the original source, to confirm that the information given by AI was correct.

Takeaway

Using AI for provenance research was an incredible tool to hunt down information that in years past would have been very difficult to access. Ultimately, the quality of the information relies on the researcher, who must possess the essential expertise to decide which questions to ask and critically assess the responses from the search. If the answers are unsatisfactory, or don't make sense the researcher must refine their queries until she arrives at a satisfactory outcome. AI is a powerful tool that should be leveraged; it enhances traditional research methods and expert judgment but does not replace them.