The Williams College Museum of Art is pushing the boundaries of curation and blurring the lines between art and technology.

By Denise Valenti

A digital collage of works found in WCMA.
A hi-resolution image of the 15,000-plus objects in WCMA’s digital collection. By Chad Weinard, Mellon digital projects manager

What is pink? It seems a straightforward enough question. Pink is flamingos and cotton candy. It’s pencil erasers and Pepto-Bismol. It can be lips or lipstick, the delicate petals of a tea rose, the firm flesh of salmon. Artists recognize pink as a tint—the hue that results when red is mixed with white.
But line up every single thing you can think of that’s pink, and its true nature becomes less apparent. It’s berry, leaning toward purple and blue. Its undertones can be orange or even yellow. Its overtones can be political or gendered. Pink is probably as complex, mysterious and symbolic a color as we can identify, making it naturally compelling as a subject for an art exhibition like the one currently on view at the Williams College Museum of Art (WCMA).

In terms of its subject matter as well as its development, “Pink Art,” which opened in September, pushes the boundaries of traditional curation. It’s based on technology developed by Professor of Computer Science Duane A. Bailey and the Office for Information Technology to “crowdsource” a definition of pink. Using that definition, Bailey and his students wrote algorithms to mine and sort WCMA’s digital collection of more than 15,000 objects, ranking them in order of their pink-ness. Curators used the results to select the pieces that now hang in the Faison Gallery.

A photo of a display from WCMA that shows the accession numbers of artwork that the museum no longer has.
Photo by Arthur Evans

The exhibition, featuring 21 works of art, is just the latest in a series of projects and innovations at WCMA aimed at “generating new models for how a museum’s digital collection can catalyze teaching and learning in the liberal arts,” says former Class of 1956 Director Christina Olsen, who recently became director of the University of Michigan Museum of Art.

Another exhibition, last spring’s “Accession Number,” considered every single work of art acquired—and in some cases lost, damaged or sold—by WCMA over the period from 1960 to 1962, based on the number it was assigned when it entered the collection. The result was a rare, unfiltered look behind the scenes at the museum.

Building on WCMA’s role as a campus museum, a new working group is investigating ways to leverage the digital collection to more thoroughly infuse art throughout the curriculum. In addition to providing gallery and classroom spaces designed specifically for the close study of objects, the museum is gathering the data and context generated when faculty and students make use of those them. That information, in turn, will help deepen the understanding of and add dimension to the collection.

WCMA is doing all of this work with the help of a three-year, $500,000 grant from the Andrew W. Mellon Foundation aimed at strengthening the development of digital resources and academic uses for the collection. In the long term, the goal is for WCMA to become a model for other campus museums by inspiring new ideas and practices. “We know the pedagogical potential of both the physical and digital collection is enormous,” Olsen says. “This project deepens and widens the value of museum collections in the 21st century, especially in higher education.”

A photo from the WCMA exhibit Pink Art. Arts with various shades of pink hang on a pink wall.
Photo by Arthur Evans

Imagine lining up a museum’s entire collection and organizing it by the predominant color in each work of art. Or sorting it chronologically, by the year each work was created, or when its artist was born or when it entered the collection. What if you could scour tens of thousands of objects and select any having to do with, say, pink?

Not very long ago, such endeavors would have been impossible. Only in the 1990s did museums start building databases of their artworks. WCMA began in 1991, with a grant from the National Endowment for the Arts, says Rachel Tassone, WCMA’s associate registrar. It took the better part of a year to enter data manually from original catalog cards and curatorial files. It took another decade for digital images to replace lm in documenting every piece. In 2009, WCMA began to fully digitize its holdings, beginning with its African collection.

Another huge shift came when museums began offering their digital collections to the public for use, says Rich Cherry, principal of the museum consulting firm Modern Operations and co-chair of the annual Museums and the Web conference. Just this past February, the Metropolitan Museum of Art made all images of its public-domain works accessible (more than 375,000), along with data for all of its digitized artworks (about 440,000 to date). WCMA’s newly redesigned website features a prominent link to its fully searchable digital collection at

Providing the public with the technology to explore and experience artworks—without the help of a curator and outside the walls of a museum—is cutting edge, Cherry says. So, too, is transparency. “This day and age is about engagement,” he says. “Not just in museum spaces, but online and in social media. Museums are inherently relevant, but there’s an opportunity to reach even more people when they’re able to engage a museum on their own terms.”

“Accession Number,” which ran at WCMA from February to July 2017, was all about transparency and engagement. It challenged traditional notions of curation and was notable for its openness about the work of museums and the life of the objects within their halls.

The exhibition focused on a pivotal time for both the museum and society at large. Under then-director S. Lane Faison Jr. ’29, the Lawrence Art museum (which later became WCMA) acquired 396 works of art, including ancient Egyptian amulets, Chinese Qing dynasty vases and a massive painting of Niagara Falls by William Morris Hunt. Each object was assigned a code—an accession number—recording both the year and sequence of its acquisition.

Two-hundred ninety items, displayed in order of their accession numbers (from 60.12 to 62.34), were included in the exhibition. Works of art that were damaged were hung anyway. Those that had gone missing or were sold were represented by square outlines on the walls. As part of the programming, local writers were asked to imagine lives for the missing objects.

“In exposing the collection’s data, we wanted to be honest, and people were fascinated by that frankness,” says WCMA curatorial assistant Jessie Sentivan. In the overview of the exhibition, curators Olsen and Kerry Bickford, MA ’17, wrote that “‘Accession Number’ poses different questions: What did the museum prioritize, why, and what did it overlook? Which works of art do we still value today? And, finally, how do we infer meaning from this or any archive?”

“Accession Number” explored this last question in a nearby installation where the public could use an iPad to curate their own digital displays of the artworks. The interface was developed by Bailey and three students—Julia Kawano ’19, Evelyn Mahon ’18 and Javier Esparza ’20, with help from the Office for Information Technology. “The content and presentation of the show defied the notion of curation,” says Sonnet Coggins, WCMA’s interim deputy director. “Not only was the content breaking convention, but the convention we used to present the content was being shifted.”

“Pink Art” shifted those conventions even further, asking the question: “What happens when you see the museum’s collection through the lens of one color?” Pink was chosen for its history, symbolism and cultural relevance—and also because it’s more challenging than one might think to define it.
Bailey and his students worked with information technology staff to develop a new app to get at that definition. The app, designed for use on an iPad, showed users blocks of color and asked them to choose which ones they considered to be pink. Hues ranging from magenta to flesh tones were interspersed with other colors, including blue and green. The app collected more than 34,000 responses that were used to define what Bailey calls “a common sense of pink.” He and four students used the definition to write five different algorithms to scan WCMA’s digital collection for pink art.

The process revealed the algorithms’ subjectivity, which is no surprise, Bailey says. “We computer scientists think of what we do as a creative process,” he says. “Algorithms can disagree, they can get things wrong. Some are beautiful and some less so, and the exhibition embraces those kinds of differences.”

The algorithms “saw” pink in many works of art that, to most viewers, didn’t look pink at all, while other pieces that were decidedly pink were overlooked, says Chad Weinard, WCMA’s Mellon digital projects manager. One work the algorithms passed over—but that certainly would have caught a curator’s eye—is a 1957 Andy Warhol print of a pink flower on a white background.
“The essence of this picture is pink, but it is a relatively small part of the overall picture,” Weinard says. “So in a connoisseur’s view, this picture is all about this color and the tenderness of this flower and the sentimentality of it all. But to the algorithms, it was mostly white. The conception of pink in a work of art ranges widely, and it’s hugely interesting.”

Some of the algorithms favored old photos that deteriorated over time, becoming a purplish-pink as the photographic paper and chemicals degraded. Others seeking tonal values close to pink honed in on sculptures. Curators used an average of the algorithms’ results to begin their checklist for the exhibition and then added and subtracted works. Each piece is accompanied by its algorithmic average.

“Part of what we learned and hoped to emphasize was the creative process of writing code and the creativity involved in computer science,” Weinard says. “We have this conception of robotic, almost impersonal, imperfect mechanisms that are working apart from any sort of human hand. And that’s not true at all. These algorithms are just as subjective, just as arbitrary, sometimes as full of errors and interesting ideas as any other kind of writing.”

Pushing the boundaries of curation is just one outgrowth of the Mellon project. The larger goal is to use data and technology to create a model for faculty, staff and others teaching and learning with artworks to contribute their knowledge to WCMA’s collection.

Museum staff regularly collaborate with professors on exhibitions, programming, installations and publications. This semester, students in anthropology professor Antonia Foias’ class The Seeds of Divinity: Exploring Precolumbian Art and Civilization in a Museum Exhibit are examining how divinity was materialized in everyday life and researching artworks from WCMA and other collections for an exhibition this spring. Students in biology professor Matt Carter’s Neural Systems and Circuits are using works by Richard Joseph Anuszkiewicz, Wassily Kandinsky and June Wayne to better understand the functional organization of the vertebrate brain.

Photo of Object Lab. One student sits to the left, writing in a note book surrounded by art work.
Photo by Arthur Evans

Artworks for these courses and seven others currently hang in the Object Lab, a hybrid gallery-classroom that WCMA describes as a “responsive pedagogical platform.” Working with Elizabeth Gallerani, curator of Mellon academic programs, faculty select objects for course-specific installations that hang side by side along with those of other classes.

Now in its fifth semester, Object Lab moved in the fall to a larger space that’s more suitable for faculty to actively teach class sessions and for students to linger and study. Gallerani also works closely with faculty to customize course sessions that take place in the Rose Object Classroom not far from the Object Lab. These sessions allow for close study of works, some of which are rarely seen on gallery walls.

WCMA is gathering more information than ever about which artworks are being used and how, in hopes of helping more faculty make even better use of the collection in the future, Weinard says.
“When a physics professor uses objects for a class in Object Lab or Rose Object Classroom, not only is the professor getting new teaching tools, but he or she is also contributing knowledge to the collection,” he says. “By providing physics context, the professor becomes a co-creator of collection knowledge.”

Though it is only the first year of the Mellon grant, it has become clear that the gathering and sharing of data has potential far beyond the classroom. Technology is showing promise to create even greater value around objects and collections, beyond their intrinsic historical, aesthetic or monetary value.

“The Mellon project essentially redefines collection knowledge by capturing new information and context generated by faculty, students and museum audiences more broadly,” Coggins says. “When this happens, so many new meanings are unlocked.”

Denise Valenti is a New Jersey-based freelance writer.

The Art of Computer Science

Why algorithms can’t agree on “pink” art.

Algorithms are sets of rules to be followed in problem-solving operations, especially by a computer. They are widely considered to be objective, impartial and even dispassionate. But as the exhibition “Pink Art,” on view at the Williams College Museum of Art (WCMA), demonstrates, algorithms are just as subjective as humans—because they’re designed by humans.

To help develop the checklist for the exhibition, computer science professor Duane A. Bailey and four students wrote five algorithms that scoured WCMA’s digital collection for pink works of art. Rarely did the calculations agree with each other. Only one identified Andy Warhol’s Flower, a depiction of a pink rose on a white field, as pink. None saw pink in John Hoyland’s Red Block on Pink, Willem de Kooning’s Springs (from “Door Cycle”) or Joseph Cornell’s untitled, mixed-media display of mostly pink rubber balls.

The artwork that did the best job of catching an algorithm’s attention was Eduardo Luigi Paolozzi’s Fifty-Nine Varieties of Paradise (from “General Dynamic F.U.N.”). But perhaps it makes sense that the algorithms viewed the distinctively pink lithograph differently than the other works. In his lifetime, Paolozzi was best known for exploring the interplay between art and technology in his work.

Here are the five algorithms and their designers:


Computes the percentage of “pink” pixels in an image using a crowdsourced
definition of pink.

Duane A. Bailey, Professor
of Computer Science


Sees images through Crayola-colored glasses and tallies the percentage of
pixels that match one of
11 Crayola “pinks.”

Lily Hyerin Lee ’17


Simplifies the image into
the eight colors that best represent it and calculates the percentage of “pink” pixels generated in the process.

Haley Lescinsky ’18


Converts an image into flat, outlined shapes, like the hand-drawn animation cels in classic cartoons, and tallies the resulting “pink” pixels.

Maria Mejia ’20


Sees the image as collections of discrete color shapes,
or islands, and then reports the percentage of shapes that are “pink.”

Jordan LaMothe ’17

Visit to see the exhibition through the lens of each algorithm.

Crowdsourced Color

To help develop the Williams College Museum of Art exhibition “Pink Art,” Computer Science Professor Duane A. Bailey set out to find a consensus for what constitutes pink. He built an app that would define the color by comparison—asking people what they think pink is and isn’t. “Crowdsource is a useful way to get a sense of what we mean about things like pink,” he says.

His app works like a game and can be somewhat addicting—so much so that it has received tens of thousands of responses. Each round offers users a choice of 16 squares. Users click on the squares they consider to be pink and submit their results. Try the app at

As if pink weren’t challenging enough to define, Bailey and his students decided to tackle Williams purple as their next project. As an Amherst graduate, Bailey says he’s particularly sensitive to the results of this exercise. “I think it will vary quite a bit, and it might intrude on Amherst purple,” he says.

You can participate in crowdsourcing Williams purple at