Oral History of Museum Computing: Marla Misunas
This oral history of museum computing is provided by Marla Misunas, and was recorded on the 22nd of February, 2021, by Paul Marty and Kathy Jones. It is shared under a Creative Commons Attribution 4.0 International license (CC-BY), which allows for unrestricted reuse provided that appropriate credit is given to the original source. For the recording of this oral history, please see https://youtu.be/qg-W8Zymrho.
In my many, many years of working in museum computing, the issue of “invisible work” and workers has come up a number of times because everybody wants to have sexy exhibitions and sexy new accessions, and most people don’t realize all of the work that goes on. It’s really hard to get funding for things like new systems and putting systems together in a network, and upgrading the network, and even original data entry. That comes up in the department I’m in — Collections Information and Access — in the Collections Exhibitions and Design Division, which is a recent reorganization. We have talked pretty regularly, but always in the abstract, about how to show the kind of work that we’re doing, because exhibitions and accessions are pretty visible, but all the stuff that goes into making it possible for them to be visible is not so well known.
I have two stories that I wanted to tell, and, and probably lots more after that, or in between. When I first started at SFMOMA in 1992, they had something they called the Collections Database, which was built on top of the museum’s financial system in DOS, and it was actually a really large doorstop.
Nobody knew how to use it. It was completely dead. It wasn’t updated or anything like that, and so the museum administration wanted the Registrar’s Office, where I was at the time, to come up with a recommendation for a system that would track the museum’s finances, membership, donors’ contributions, the art collection, and, oh yeah, attendance, too. So we spent some time looking for that magical unicorn system. And, guess what? We never found it.
We worked with the company that became Gallery Systems. We were one of 20 museum partners working with them to create the best museum system, including images — hugely innovative at the time — in the early ‘90s. In 1994, we moved out of our 1930s building and into a completely new, purpose-built museum. The new database wasn’t ready yet, so we had something cobbled together that we used for data movements and inventory.
I backed that system up onto my own desktop computer every night, which was what people were doing at that time. I had previously worked at another museum where the Head Registrar took her computer with the disk and the hard drive home in her car every weekend to make sure that it stayed safe. Our building in Golden Gate Park was not earthquake safe, and it hadn’t been long since we’d had a big earthquake in the Bay Area (1989); so it made sense to take the whole database home to keep it safe. It sounds really weird now, but that was what you did at that time.
SFMOMA’s system was just the kernel of this interim database with object locations and bare bones object information, put together out of a number of different departments’ FileMaker databases — you can imagine what that was like. Every department had their own set of records, meaning they could alter their own records without bothering anybody else’s records, which is bad for data integrity. If the Publications Department wanted to put out an annual report, they would ask The Director’s office for object information, they’d ask Curator A, B, and C, and then, “Let’s ask Registration to do it.” More than one object record for the same thing had slightly different titles, slightly different credit lines, and it was obvious we really needed a centralized database. We had roughly 15,000 things in the collection, and all their corresponding paper records.
15,000 things translated to 15,000 object files in hundreds of binders. We didn’t have the staff to input everything into our new system, and the new system wasn’t ready. The museum administration decided the way to get this done was to have a third-party service microfilm all of the binders. They boxed up the microfilm in crates, and send it to a company in Australia that had won the project, being the lowest bidder. Because that’s what we museums would often do, knowing full well that the cheapest thing might not be the best thing.
To make a long story short, the company went bankrupt in the middle of our project, and we had crates and crates of microfilm in Australia, where probably most of the records had been entered on a first pass into something like FileMaker. […] No disrespect to the people entering the records, but they were not familiar with our terminology, and they may have had a hard time reading the sometimes confusing handwritten records written by people in a hurry. We certainly did, and we were familiar with the language. Because of that, the staff in Australia would have been expected to interpret what they read without any prior knowledge, context, or any consultation with us.
We were desperate to get this material back in whatever condition we could, and we knew it was going to take a long time to clean up. It took us a full year to clean up everything. The records got into the database, but it took a long time, and it took another long time to get the different departments — Curatorial and Registration and Conservation and Exhibitions — to not only trust the data, but actually use the database.
It was at least a five-year rollout, and not everybody used it even at the end of that period. Some people thought it was unreliable — it feels like someone is criticizing your child when they say that. I had to learn not to take it personally! Whoever is spearheading the project is going to get all the blame regardless of whether the trouble is in the data or the functionality.
[Marty]: Let me ask a quick follow up question. In this period, then you’ve got people modifying paper records and the database records, at the same time, so they’re diverging?
To be realistic, the paper records didn’t really get changed that much because the data just didn’t change that much. These days, the digital records are the most current ones and people sometimes forget to update paper records, so we compare the two sets before we put things online. Newer paper files are usually lighter, so comparisons can be faster.
And there are always people in the institution who won’t use the database no matter what you tell them, or how many times you train them.
A nice thing about our EMPs [Emerging Museum Professionals] is that museum studies students know how to use databases, so as these new professionals were filtering into the institution, they adapted to our new system quickly. We had one director (years ago) who wanted the entire database of object records printed on paper and shipped to his home, you know. [Laughing] And, we did it!
[Marty]: Just you, you made a neuron fire in my head. When I was a doctoral student at University of Illinois working at the museum there, we had a professor … I forget… I want to say anthropology, maybe, come to the museum, to look at something and I showed her how to pull up the records on the database and she looked like I’d slapped her, and she said, “I didn’t get a PhD to type on the computer,” and she made me print them out! Okay!
Some of our projects pushed us along and modernized us a bit. SFMOMA was part of the MESL Project, (Museum Educational Site Licensing), and later we were part of the AMICO Project (Art Museum Image Consortium), in the late ‘90s, early 2000s. At the same time, museums were putting up very simple websites. Most places didn’t have online collections, but a few of them had early attempts. As soon as those other museums started putting their collections online, SFMOMA trustees wanted to see our collections online, especially with things with their names in the credit line, you know — I totally get that. It took some years to get the records standardized and ready to publish. To say nothing of the images — in a modern art collection, the number of images that you don’t hold the copyright to far outweighs the ones where you can claim public domain.
It seemed like suddenly, it was much easier to find things, and we had to start paying more attention to getting permission to publish images of artists’ work online. I hate to say it, being in the older crowd myself now, but a lot of the older artists didn’t want their images of their work online back then because they thought it would cheapen their artwork and make the actual object less impressive. Of course, it isn’t true and we have a lot of data on that. It reminds me of that old argument, “We can’t put color plates in books because then no one wants to see the objects.” We know that’s not true.
That kind of attitude was something we had to break through and get beyond, so some of us were pretty excited about the AMICO Project. We had to submit data for the world to see and we had to submit images, meaning we had to scan our transparencies. And, sometimes take new pictures of work, because some of those old transparencies weren’t always good enough. We were scanning transparencies and people started seeing images in our internal database, which was pretty exciting. But we didn’t have transparencies of everything, so we didn’t have images of everything. We did have Polaroids from all that inventory we had done before we moved into the new building. We started scanning Polaroids and popping those into the database, so you would see a beautiful picture of a Matisse painting scanned from a high-quality transparency and the next record would have a picture of something on a rack with plastic around it. Even so, people began to depend on images — it’s so much easier to find something if you know what it looks like.
Eventually, having images in the database became the direction that everybody was going in. We had the Steve crowdsourcing project on the horizon, and AMICO, which was important in terms of its proof of concept. It didn’t have an exceptionally long life but, but it really proved what could be done and how people really wanted that kind of information. Now, we are publishing things to Artsy [artsy.net], and it’s much easier to submit your material.
Steve was a public crowd-sourced tagging project that a lot of different museums worked with — LACMA, for example — unfortunately, we were never able to hook the Steve widget up to our system [so we didn’t use Steve with our website; but we were part of the discovery phase]. Still, the Steve project proved that crowdsourcing tags was really helpful, despite the seventh graders who put things like, “I really like Jeremy. Why doesn’t he like me?” in the tags. Internally there was discussion about: Who’s the audience? Are we getting our art historical experts to tag the objects? I think some museums were able to interest their curators in the experiment, but ours had other priorities. They felt their peers didn’t need tags.
Folks in the Collections Information and Access area, where I am, knew that it would eventually be helpful to have tags that a general audience could relate to. The CIA team began tagging objects with subject tags and came up with some guidelines, based on The Fine Arts Museums in San Francisco’s system. It wasn’t a structured hierarchy like some places used back then. Staff at SFMOMA didn’t like hierarchical vocabulary and wouldn’t use it.
Fine Arts called their approach “Word Soup.” They had been uploading their not-for-publication images into their database, and folks took their work computers home with them to do tagging at home. They tagged things they noticed in the pictures, regardless of hierarchy. Which kind of goes back to what we were talking about earlier, working at home. “That is so great! You know? We could do that too.”
The working-at-home part didn’t happen, but as part of our process of reviewing the records against the paper documents, we put subject tags into each record that we were approving in the database. We were hoping someday those objects would go on the website, and that people would be able to find them, using the terms that we added.
Over 40 percent of the collection is on the website now – 40 percent of 50,000 is 20,000; it’s more than that. All of those records have tags that were entered by me and one other person, mostly. We had help from other departments occasionally and from volunteers and interns, but so far as the steady application of effort, that came from the database team. About five years ago, Layna White, the Department Head for Collections Information Access (now the Division Head), thought we should start looking for bias in our tags. The other tagger and I were both white ladies of a certain age with master’s degrees in museum studies, so we knew that there was bound to be bias in our tags. We tended to not only tag the major things in the image, but we’d tag things that we thought were interesting. We would tag “man” and “suit” and “hat” and “necktie” and “spectacles” (which no one ever looks for) and “glasses.” We also tagged things we thought were interesting that weren’t the main focus of the image. I’m a dog lover, and so I always tagged “dogs,” and “animals.” If I knew what breed the dog was, I would add that too. We definitely had our own biases, and so did the institution.
Biases and technical issues came into focus when we opened our new wing in 2016. There was a lot of hoopla around the new building, and everybody was really excited, the public had been hearing about it for a long time, and people wanted to see themselves reflected in the collections that we had in the galleries and online. We had to change our approach to tagging with the input of visitors and the people who were contacting us through the website.
One correspondent emailed me and said, “I’m looking for images that I can show my kids of Black people represented in museum collections. How do I find that on your website?”
Right around that time, we also had a lot of interest from people who wanted to see the collections that weren’t on view in the galleries. Our digital technologist, Jay Mollica, came up with a texting service called “Send Me SFMOMA” which blew up the Internet and was amazing, incredible, and was recently sunsetted, sadly.
I will go to my grave wishing “Send Me” was still available because the awesome thing about “Send Me SFMOMA” was that you didn’t need a computer, you just needed a phone. You could text “Send me (something)” to a number, and the service would go through and pull out an image — especially selected for you — based on whatever word or words you entered. The person who had wanted to see representations of Black people on our website also searched “Send Me SFMOMA” for Black people, and what she didn’t know and a lot of us didn’t know was that “Send Me SFMOMA” could only search for one word at a time. It couldn’t find two words separated by a space. And so, when she searched for Black people on “Send Me SFMOMA,” she got a picture of the United Nations’ first meeting in the Veteran’s Building (where our museum used to be) in 1945. So, there was this conference room full of white men in black suits. And she thought, “Well that’s interesting. It’s not what I wanted.”
So that was a fail, and we needed to work on that. One of the reasons the person’s website search failed was that staff had never tagged or otherwise recorded what they might perceive to be the ethnicity of people in images (how would we know how people in pictures would identify themselves), and where that became a problem was the public wanted to find images based on someone’s appearance. We hadn’t known people would search that way.
So we went back and tagged people, white people and people of color, but the other white lady and I didn’t feel comfortable saying, “This is a Black person, and this is an African person.” So there were a lot of uncertainties. Some artists specifically make pictures of African American people, and we felt comfortable tagging those, if we knew the artist’s intent.
We always wanted to have more voices in the tagging conversation than just the two ladies, but it’s been really hard to get staff to add tags because usually they’re not the ones searching. We’re getting statistics about what the public searches for, which helps make the case for getting more voices. I feel pretty qualified identifying a person in an artwork as white — and I might be wrong — but that’s about it. That’s where feeling qualified stops.
One of the gifts that we’ve had with the COVID shutdown is that we have frontline staff helping out on various different projects, and now they’re starting to help us tag objects that are going online or are already online. Literally it’s just started this week, so I have great hopes that we can spread the news about doing this. Everybody who has a user account in the database can add tags, but most of them don’t know that.
In a training session last week, somebody asked me, “What terms are you using to tag works by gay artists or members of the LGBTQ community?” And, we have not gone down that path yet, because even if it’s public knowledge that an artist identifies as LGBTQ, that doesn’t mean that their artwork shows an LGBTQ theme (and who’s to say what that might be), or it could be that they think all of their artwork does or doesn’t, so we’re working with curators on that.
People are asking, “Where is your work by Black artists? Where is your work by Asian artists?” And we are starting to use those distinctions in our internal database. But we’re a long way from doing something like the Smithsonian just came out with, a feature on art by Black artists in their collection. We’re far away from doing that, and we would only do it if the artists agreed. In the meantime, we are starting to ask living artists how they want to be identified, if they are willing to answer. We also need to research artists who are no longer with us. And in some cases ULAN (Union List of Artist Names) collects that information (ULAN puts “LGBTQ” in the Nationality field) and sometimes ULAN is wrong.
We have 6,000 artists for which the ethnicity has not been recorded, and after a long time working with the historic collection, I’m going to say most of those artists were white. There’s definitely a good Latin American collection, and a pretty decent Asian collection, but honestly, if you assume most of the 6,000 people in the historic collection were white, most likely your assumption is going to be pretty doggone close. That’s my opinion, and it is not popular.
We are now intentionally collecting artwork by more diverse kinds of people than just white male artists, and I’m hopeful that direction continues; it’s in our strategic plan, and it’s something that we’re all working on. I hope that in 10 years, our collection and exhibitions continue to be more varied. That said, we have a lot of work to do! And we’re working with curators. They are coming up with a questionnaire for living artists [the first one went out in 2021] and there are other sources of that data that we’re also trying to hit, like Wikidata, SMU, and ULAN, for deceased artists. But, honestly, finding a record that identifies someone as a white person or a person of color is very difficult. We’ve been running test sets trying to find that data, unsuccessfully. When you find things, it’s great, but the work where you don’t find anything is also interesting. As an industry, we have spent such a very long time collecting white artists that our records have only specified ethnicity (perceived or researched) when the artist was a person of color. That is my opinion and assumption, and it’s not something people want to hear.
Still we need to push forward to greater representation, and hopefully as museums, we will be able to do that.
[Marty]: Well, I love the example you’re sharing about the, the inherent biases here, and the, and the tagging and the problems that this causes. The “Send Me SFMOMA” example is fantastic. And what it reminded me of was the problem, a couple of years ago, with a Google Arts and Culture plugin, where you took a picture of yourself and then it matched you with a painting.
Yes, thank you, that was one of those funny stories I was going to mention. We tried a bunch of different things to possibly automate tagging and we worked with Google Art Project. An Artist in Residence came up with an algorithm whereby they fed the computer images, and since we didn’t know how to tag our abstract things, we gave them 20 pictures of abstract paintings. They fed our images into the computer, and it came up with poems based on each image, which were hilarious, and completely unusable, and not something we would ever even put on the wall. The computer was trying to match the shapes it found an abstract painting with an actual thing, and so we have a Rothko painting that has this kind of horseshoe-shaped form in it that the computer decided was a sewing machine. And I can’t see it any other way now. It’s like, oh yeah! That’s the one with the sewing machine! Or it’s the one with the fire hydrant! So, that told us that we’re not really there yet, and we need to keep having humans do tagging, especially with abstract work.
When I taught collections management, I used to tell my students that two things: yes, collections management can be a little dry and you will get through this; and it takes a steady application of effort. I apply that to my own work — you have to clean the stable before they let you ride the racehorse. Even after you ride the horse, you still have to clean the stable.
And the other thing that I still tell interns to this day, is that just because you learn something in school, doesn’t mean that’s the way it’s done in whatever museum you land in. I know firsthand because I went into a registration position with a little tiny university collection, and I thought “Oh my God! Everything’s a mess! You did everything wrong, blah, blah.” But I realized, much to my embarrassment, that there probably was a reason why people set things up the way they did, and that people who have been there a while might actually be able to explain. So, that’s something that I try to tell interns in a nice way when they tell me, “You’re doing it all wrong.” Yeah, I know, I’ve been there. I have been that person.