With Museum of Things for People (MoTfP) Design Museum Gent examined the conduct and expectations of visitors in the exhibition Object Stories, which shows 200 collection pieces. The project focused on three main aspects, namely wayfinding (in what way do people move around in the exhibition?), tracking (which objects attract people the longest?) and data enrichment (suggesting points of interest in the city based on the results of the tracking). In the presentation we explain about the cause of the project, the technical realization in collaboration with two startups, and our most important findings.
Innovation Lead – Digipolis
City of Ghent
Eva van Regenmortel
Design Museum Gent
This presentation was filmed at the MuseumNext Digital Summit in Autumn 2019.
Eva: Hi everyone. I’m a bit old school I’m afraid, I need my text with me. Museum of Things for People was an experiment on using tracking technology to investigate how visitors behave in our collection exhibition object stories, and also on mediating our collection through data enrichment. Pieter-Jan will tell you about the overall choices and learnings of the project and I will explain about the reason we did this, the specific context of object stories and the data enrichment.
The reason we did this was because we’re building a new wing that should be finished in ’23. And we’d like that new wing to function as a third place, which we see as a zone between the existing museum floors and the city of Ghent with all of its inhabitants and its visitors. And this wing will have free access and it will host small short-term events that we’ll organise together with the citizens. And secondly, we want to offer them a meaningful and accessible acquaintance with the collection.
Museum of Things for People was a test case to learn about how visitors interact with objects. If tracking technology can be useful for that and if people would be interested in recommended points of interests in the city, based on the tracking. Design Museum Ghent is a small museum. We have some 25,000 pieces. Object stories shows 200 of them, not in a chronological way, but partly typological as you can see here with the chairs. And partly in the form of artistic installations in the middle of the exhibition, where objects are presented according to the personal view of an external curator.
This is how that looks on a floor plan. This is where you enter, these are the installations, the artistic installations in the middle and then the objects grouped typologically around them. You can just walk around and enjoy it or not, or you can choose one of five themes and then you have to enter this corridor, which looks like this. And you choose a theme, you take a booklet, you walk back to the exhibition and you read 10 stories about 10 specific collection pieces and how they fit into the theme that you chose, which is all about different ways to look at a designed object.
Now for this Museum of Things for People project, we invited visitors to wear a tag around their neck. And then we measured with Ultra Wide Band positioning technology, how much time they spend in front of specific objects and in what way they circulated in the exhibition. And as they leave the museum and fill in the number of their tag on a display, we tell them how much time they spent in the exhibition, how they moved around, what object they were near the longest, and then we call it their favourite object. And then we offer them three recommendations to points of interests outside the museum, based on that favourite item.
And then they can like or dislike these recommendations, which is over here with the thumb up or thumb down icon. I personally believe these recommendations make a strong concept, especially for design collections, because leading visitors from a piece of material culture to places in the city is what brings objects alive, keeps them relevant. And in their turn, the objects add value to the cultural assets in the city.
At first, the intention was to automate these recommendations by matching them with existing external datasets. But unfortunately, there weren’t many available. An interesting one was this impressive database of the built heritage in Flanders, which is an instrument for the Flemish Government to enlist all valuable buildings. We tested if we could match our collection objects to historical buildings in Ghent, based on their stylistic features. But the result was disappointing, because the database of the Flemish Government is an analytical instrument. The descriptor postmodernism for instance, is also allocated to buildings that only show minor post-modern details and we needed the convincing ones. And making matches based on the dates of the objects and buildings gave the same rough results. The connections weren’t sharp enough. So you still would need to select manually the relevant ones.
The database of the public library was a better match. Again, based on stylistic features of objects, we linked them to the most recent book in the library that had the name of the style, like [inaudible 00:05:55], modernism postmodernism in its title. So we ended up using the existing database of the public library and making two other curated recommendations ourselves, based on our own knowledge and imagination.
I was involved as a researcher in the collection department to make those curators’ recommendations. For instance, for Henry van de Velde’s piano bench, we referred to The Book Tower, which is his only building in Ghent but one of his most famous. And on the top right, to a far less known restaurant, near to the city centre, where clients can eat sitting on benches that Van der Velde once designed for the Belgian Railway Company.
And these are some of the other curated recommendations. Some of them are historical buildings, others are galleries or even shops. Some of them have a rather associative link with a collection item. For example, How High the Moon, which is the name of a chair by Japanese designer Shiro Kuramata, but also a famous jazz standard. We refer to Ella Fitzgerald and to the Hot Club de Gand, which is a jazz club in Ghent. The results of all these measurements with the tags were processed and presented to the museum staff in a dashboard and from the 13th of June to the 19th of October, 1,423 visitors volunteered to wear a tag to help us doing this, which is nice.
And for me, it was interesting to see how 898 of them made it to the corridor, which are these. Made it to the corridor with the themes and the booklets. But that means around 500 of them either didn’t get the full concept or weren’t interested in the themes. Apparently curiosity, which is the red one was the most popular theme, which is about objects that in some way or another seek attention more than others and act as conversation pieces. 940 people of those 1,423 filled in the number of their tag in the display and got to see their favourites objects and the recommendations.
What was a bit disappointing to me making those recommendations was that only very few people got tempted after that to respond to those recommendations with a thumb up or a thumb down. Together, these 900 people only liked some 80 recommendations and disliked 20 of them. So for me, in the end, I didn’t learn if visitors appreciate being guided to points of interest outside the museum. And I guess we’ll always need to keep asking people in person about their experiences and expectations. But in general, the technological side of the project was useful because the tracking system worked and Pieter-Jan will now tell you about that technological part of the project.
Pieter-Jan: Okay, thank you, Eva. So just to give some heads up about why we did this project and what it actually does with the tracking, I just wanted to present this very briefly. I’m Pieter-Jan, I don’t work for the museum. I work for Digipolis, we’re the IT government agency for the City of Ghent and specifically in innovation lead or innovation manager, if you would like to call it that. So I work both for the green service, fire department or a museum as well, so it’s not just about this.
And the question I got asked by Katrien Laporte, the director of the museum was, “Can we have more insight in how people behave in our museum and their specific interests?” Because now we know stuff from the Google reviews online, we know stuff from the sales of the tickets, but we’re not so sure what they do inside of our building and what they actually like in the first place. So we searched around for innovation budgets and we found the Smart City grants in Flanders for Flanders Innovation and Entrepreneurship. We submitted a proposal and we got it and we had a year long, the time, to experiment with IoT technology within the museum.
And we specifically focused on indoor positioning technology to really see, for example for this lady, at what objects is she standing still, and is there actually a sort of interest that we can derive from this data? And it wasn’t just with a tech focus, just let’s hang sensors all over the place, but also think about what does this mean specifically for people? What is the value for the visitors in a museum and also what does this mean for their privacy in the first place?
And aside from hanging up the Ultra Wide Band, that Eva mentioned, we also did a lot of desk research and what technologies are available now on the market that we can use in order to do in their positioning in the museum? So we chose the top right corner, of course, because we really looked at the accuracy of the technology. Can I really detect how close by I am towards a certain object? But also, the maturity specifically for an indoor museum. So that’s both, is the pricing okay? Is it mature enough as a technology to go forward? And also, what does this mean in terms of the overhead for the customer, what do they need to do to start with that technology? And also, the setup overheads, what does this mean for the ticket office?
And just a good example for that, one of the most accurate technologies in positioning is motion tracking, but we’re never going to convince a visitor to wear a funny suit to be tracked. And you laugh now, but it’s is the same, for example, with an application. Just try to imagine your visitor to the Hex Museum, for example, and you walk in and you have to say your age category in order to know the pricing of your tickets. And you need to say where you’re coming from for the statistics. And then the ticket office also says, “Yeah, we have an app, please take out your smartphone and press this on the Google Play store. And no, no, that’s not a real app, this is a real app. And okay, you have a Windows phone. It’s not going to work or there’s no wifi or whatever.” There’s too much overheads.
And what we actually did, is create this little wearable thing on a lanyard and the only thing that we said to people at the ticket office is, “You can wear this around your neck. Would you like to, yes or no?” And based on that, we can give you guidance towards other places in the city based on your interests. So that’s the only thing that people needed to do.
How does it work? Inside of it, the little green thing is a radio frequency system that connects with anchors, we call them, in the ceiling and they are able to triangulate the position of that specific person. It’s technology from Posyx which is a spinoff from the University of Ghent. And so, on an accuracy of 10 centimetres give about, we know where people are in the museum. And it also has a very low latency. That means that we get a lot of data, around six calls per seconds, about the position of a certain person.
So the sensors in the ceiling send that to the positioning server and they send it then back to the clouds and the data looks like this. So it’s a lot of XYZ coordinates from a certain sensor that comes out of it, but you still need a lot of data analysts to translate this into interpreted data. So we also translate this to, for example, a routing. So you see the museum a bit tilted from the former example, but we also use for example, clustering analysis, to know where somebody has been standing still for quite a while. And we try to use that to calculate the interests of a certain person.
So we put that into a computer that’s smarter than us to calculate those interests and send it back to a user interface that people can use at the end of their visit. And once they’re done with that, that data is being written away, so it’s anonymous data, and we use that anonymous data to pull it back via Google BigQuery in order to make a dashboard for the museum to use for their statistics. So we’ve already shown this, so it’s about your favourite objects, your favourite style, and where can we guide you towards in the city, whether it’s a building, a book, which is based on the API of the public library. So it’s not just, we’re referring you to a book, but it’s also the book that’s specifically available in The Krook library.
And afterwards, we also show this, we just give them some insights in their own data. We started this as a way of providing a sort of quantified self. I can measure my own habits in daily life, but it’s also about just giving some transparency on what does the algorithm know about you? What does the museum know about you in the first place and what do we do with that data?
And now, this was a demonstrator specifically for guidance towards other places in the city, but you shouldn’t use inner positioning just for that. You can use it for way more other cool stuff. We actually had 18 use cases, mainly based on people prototyping, but for example, these are the three big ones on the upper hand. For example, way finding. When somebody is heading towards an exit, that the museum can say through screens or a voice interface, whatever, “Maybe you should stop. You missed a whole floor.” Or, “You missed your favourite object. Maybe you should turn back and watch that as well.”
But it’s also about information and experience. When we did the user research, people said, “We don’t want AI to guide us through a certain route in a museum. We want to see everything, but we would like to have some input on the scale for example.” People that are not seeing well, especially old people are like, “Object texts are too small. Can’t you just give me an [inaudible 00:16:18] screen that just enlarges whenever I’m passing by.” So you could do that perfectly with this.
But you could also use it for example, to have a very dedicated audio guide. For example, if I have a very fast pace, which I have in a museum, I don’t walk around quite slowly, that it just gives the basic information about the exhibition. And if you have a very slow pace that it gives them more information about the specific objects you’re standing near to. And for the museum, we’ve shown this, but I’m going to just show the other one. It also gives some insights on where are people standing still, specifically also for the thematic rooms and the objects where people stand still near to, but even for, and that’s like the little red squares, the objects text, are people actually reading those are not? So that’s also something we can calculate because we’re that accurate.
But there’s a cost hanging to it, of course, like every time. So based on the projections on the budget that we had per square inch, this is about 41 euros to set this system up. And it’s also about 10 euros per square metres to instal this, but take into account, we installed this technology into the ceiling while the exhibition was live, so we had to be careful with ladders and other stuff, not to bump over certain objects. And that’s just the tip of the iceberg. It’s also about what’s happening behind. So it’s also about maintenance of that hardware. It’s about support. It’s about setting up servers and databases and calculation of those models. So you need to take into account at least for a thousand square metres that you have a cost of around 15,000 per year.
But it all depends on what you actually want to do. How accurate does your data need to be? I think for data on a museum level, the most people already know that that’s your ticket sales and taking into account people who went by free, for example, for a business event. But if you want to look, for example, to data on a thematic level or an exhibition level, you could, for example, use cheaper technology like Bluetooth or RFID tags with certain windows where you pass by and the system knows somebody went just in that room. But if you want to have data on an object level, you at least need this kind of accuracy in order to have that calculation.
So that was about it. Because this was a project based on public funding, it’s also going to be about public findings. I’m wearing a Creative Commons T-shirt as we speak. So everything from how the algorithm was based up, our user research and so forth, will be published on MOTFP.BE. Apologies, it we’ll be in Dutch though, but feel free to check that out. Thank you.
Sarah: Thank you. Eva, please join us on stage as well. And thank you also, Pieter-Jan. We have a couple of questions here in Slido and the first one and most popular one is, to what extent do you think the visitors awareness of them being tracked, influenced their behaviour in the exhibition?
Pieter-Jan: It affects it definitely. Definitely. Definitely. And maybe a bit too much, because for example, we also could have gone with, for example, facial recognition with cameras, and that’s almost an invisible technology. It does it anyway, without you opting in or out. The fact that we use this technology will make people perfectly clear what we’re actually doing and they also have the opportunity to just say no, and I don’t want to wear it in the first place.
And that affects the behaviour, but people are very keen on it because they say, “Okay, the museum is part of the public domain, so there needs to be awareness about what’s being done with my data.” And in that way, at least they’re sure about what I’m doing, but of course it affects their behaviour a bit.
Sarah: Yeah, do you have something to add to that?
Eva: No, not really.
Sarah: Do you get issues suggesting commercial places like shops or restaurants accused of hidden advertising or using public money or unfair choosing, so where to direct them after they visited the museum?
Eva: You mean if it’s like a problem that you direct them to commercial places as well?
Eva: Yeah. They’re not really, really commercial… They’re commercial, of course, but they always had some link with the object shown in the museum. We’re a design museum. We’re all about daily utensils, furniture and so on, so there’s an evident link to shops in the city. It wasn’t all about shops. Most of the recommendations went to historical buildings, but we thought of varying them a bit. And especially also to see if people would like the commercial recommendations.
Sarah: Yeah. And I mean, design is also part of the creative industry. So, I mean, there is economic-
Sarah: Yes, okay, understand. As a museum, to what extent do you think recommendations should introduce people to things they don’t already know they like? [inaudible 00:21:14] Did you ask them if they were surprised or do you know if that happens?
Eva: Well, they should have responded to the recommendations and not many people were feeling like doing that. So we don’t know if they-
Pieter-Jan: I think the biggest challenge there is, we don’t want to let them feel overpowered with having to fill in three form fields about what they felt about the technology, how they went forward with it. Did they actually went to that exhibition that we proposed and so forth?
So we’re still thinking about how to capture that qualitative data as well, because a lot of the people who work in the museum security ticket office pick up a lot of that data, but it’s not being transmitted in a structured way in a sort of document in order to do something with that. But it’s something that we’re really struggling with and it’s still a challenge.
Sarah: Well, and it could be a nice surprise also to find out one day, it’s still something to wish for. Let me see if there’s another question. Do you have ideas on how you could make this project more successful and to get people more interested in your recommendations?
Eva: I think one of the explanations could be that the display was kind of small. They had their favourite objects quite big and then the three recommendations rather small. So people had to have a certain focus and really need or want to have that information. Maybe the display could be organised in another way.
Pieter-Jan: Especially, and this is something that we didn’t take enough into account when we rolled this out, is people don’t just automatically start using technology. You really need to advertise that in a way and it was also a struggle with the design museum who said, “You’re not going to hang my ticket office full with posters about your technology.” So that’s a conversation you need to have.
Sarah: Fortunately you can’t, no.
Pieter-Jan: But it’s something that really needs to grow. I think the biggest issue that we have as a museum is because it’s about recommendations towards that city. And of course, a lot of the museums in Ghent have a big link to the city as well, is that we really need to think about the collection data and how would this all link together so that we make more interesting connections towards other places and towards other artworks within the other museums in the city. And I think that’s going to be the biggest challenge, how to link those collections with each other, so that we can automate in a much more structured way, yeah.
Sarah: Interesting. And a question actually from me, did you also think about how to make a guided tour and these recommendations kind of collaborate, like there is someone who’s walking towards you and saying, “Hey, I know a place where you could go.” Does it work? I don’t know, talking about this physical digital space.
Eva: That might work, but that would be-
Sarah: Would you try?
Eva: No, no, not yet. But maybe someday.
Sarah: Okay. All right, thank you so much Eva and Pieter-Jan. Let’s hear it for them. Great.