Over the last six months an interdisciplinary team of experts in Science and Technology Studies led by Daniel López from Universitat Oberta de Catalunya’s CareNet (Care and Preparedness in the Network Society) has canvassed an ‘extended peer’ network for Module 3 of CANDID, addressing some of the most contentious issues arising from the operation of ‘sensing infrastructures’.
Having explored the controversies, these Social Science and Humanities (SSH) scholars highlighted three main areas of concern.
Firstly, unintended effects coils arise relating to people’s perception and level of awareness of aspects of the complex related problems and issues. Sensing may, for instance, contribute to reduced attention toward non-quantifiable aspects.
The second issue relates to the suggested opportunities offered by sensing infrastructures for empowering citizens. Policy makers and planning authorities looking to implement ‘smart cities’ tend to emphasise such potential but public participation programmes tend not to flourish within top-down governance models.
The third area concerns the ethical and practical implications of machine learning used to analyse the large volumes of information gathered by sensors. Such big data analytics tools are intended to make sense of this data, but the algorithms used as the building blocks for interpreting it all can replicate discriminatory practices, can be based on faulty assumptions about signs of anomalous or abnormal behaviour, can introduce mistakes and can be provided with inaccurate data. In addition, the more systems are automated, the more should be asked about who should pay for errors made by such autonomous machines.
So, how can the CANDID project start to explain these observations, and what sort of alternative vision could help enable the Social Sciences and Humanities to constructively engage with innovators?
Sensor Infrastructures
Installing sensor technologies, within geographical regions, organisations, and even people’s bodies can enable systems to be built with capabilities for learning and reacting to changes in physical environments.
Sensor technologies can be used to measure physical attributes such as light, heat, motion, moisture, and pressure, for conversion into data and eventual interpretation by a human or artificial observer (such an actuator or warning systems).
Interconnections between sensors and the systems within which they are embedded can enable sentient, or sensing, infrastructures, with claimed transformational abilities for urban design and for how spaces are conceived and experienced.
Sensor infrastructures enable such use cases as:
- RFID sensors in retail and the apparel sectors, for monitoring goods along the global supply chain
- Biometrics, for identifying travellers as they move across borders
- Facial recognition software, for identifying suspects and triggering alarms for anomalous behaviour
- Information from mobile phones, social media, weather radars, earthquake monitoring and seismic systems, merged and analysed, for anticipating disasters and communications
- Within smart city projects: monitoring pollution, noise, traffic, energy consumption, transport use, as well as citizens’ connected devices
Responsibility to explain
Coming from the multi-disciplinary social science field of science and technology studies, the CareNet researcher Maxigas said that one of the things that sets his field apart is that, more than simply observing technological developments, is that it also attempts to explain technology itself.
“We can’t simply use the rise of Information and Communication Technologies in the last part of the twentieth century to explain social phenomena. We have to explain the rise of ICTs too.”
“This is called a social constructivist framework. We don’t take anything for granted but we try to explain why things happen and how society is shaped.”
“Sensors are usually seen as solely passive. Roboticists, for instance, use sensors for inputs and actuators for outputs. They see sensors as gathering data about the environment – but we show that even just by gathering data they also change the reality around them.”
Preliminary observations – visibility and ‘build it and they will come’ cowboys
Although he frames these as so far just ‘haphazard ideas’, Maxigas sees some topics emerging from the responses analysed so far.
“The first problematic we identified is awareness; or how much people actually know they’re surrounded by sensors, as well as how sensors can work, the role of citizens, and how the machines fit into existing structures and social realities”, he said – ahead of a consortium meeting bringing together participants in the CANDID project, set for 26 September in Barcelona.
“One of the issues that emerges is around the invisibility of sensors. Especially in the case of smart cities, sensors are employed in urban environments, but a smart lamppost, for instance, does not look like a computer or a surveillance camera. It may not be evident that there are different aspects of people’s lives being recorded, and, even if they do realise, it is even harder to work out what happens with that data and who has responsibility for it, and what is the point behind gathering this data.”
Another aspect highlighted was that the generation of data can be seen an end in itself, meaning that maybe only during a project is it worked out how data could be used and for what purposes.
“The more pragmatic interviewees raised questions because such ways of thinking are really technology driven and diverge from norms of deliberation and policymaking.”
”Under proper democratic procedures sensor infrastructures should be approached from a data protection perspective. It should always be clear why data is gathered, what purpose it serves and who can access it and under what conditions.”
“This is how data can end up in an open data portal produced with the expectations that companies and individuals will use the data, which is a somewhat ‘Wild-West’ approach to innovation.”
Reproducing the same way of doing things
One of the aspects that emerged was what Maxigas called the ‘dark mirror’ of sensors.
“One of our initial assumptions was that sensors were for gathering new information or new input factors. Instead of sensors being some sort of sixth sense we discovered that different actors are actually using sensors for very different purposes. When gathering data it’s not necessarily the case that they want to learn something new, or if they want to, then sometimes a big investment in a sensing infrastructure simply leads to learning something that is already known.”
“One example is the Barcelona City Council, which is visibly involved in smart city developments. One of the first things they did (in a project) was place sensors in selected garbage bins to see what times would be best to empty these bins. These sensors provided quasi-real-time data to measure how full bins were several times a day.”
“As soon as someone put a pizza box in the container it was indicated as full, so they had to install weight sensors for more accurate data. After some months they obtained a small improvement in the routes of the trucks but once this one-time dividend paid off the real-time data didn’t lead to any further efficiencies or economic benefits.”
“In fact, no matter how much data they gathered they only showed the same improvement as by people managing trash collection operations elsewhere.”
“We have to ask if it was really worth it, what was the actual problem they were trying to solve, how much measurement is needed, and what kind of information is produced.”
“It seems they had just been in competition with other cities and putting the technology in the urban environment was seen as a good thing in itself.”
“Another case looked at was Making Sense, a bottom-up project where citizens built their own sensors and used these to learn about their environment.”
“In one part of this project they installed noise sensor infrastructures in Barcelona and of course what they learned from this was that Barcelona is a very noisy city, and especially in the squares people would talk very loudly.”
“In fact, the city already had industrial quality sensors installed for air quality measurements that recorded noise too so the same data had already been shared by the local government.”
“It doesn’t mean in this case that the sensing infrastructure served no purpose. It simply questions what an infrastructure system can do, because before the Making Sense project it was very hard to dramatise the issue of noise that residents may think is a real problem. By working with Making Sense citizens started to think about the question and working with the technology to dramatise the issue and present their own data to the government.”
“It seems the technology itself did not produce new information in this case but was used as an organising device, and then gave rhetorical strength to the claims made by these citizens.”
“A more straightforward case study was a project called IMVEC which worked with activists engaged in preventing illegal landfill dumping. They installed cameras to gather evidence for use in court.”
“Again, they weren’t gathering new information, but it was used as documentation.”
A Larger Theoretical Framework
The Module 3 researchers devised ‘smartness regimes’ as an experimental framework, and are continuing their analysis.
“In the written evidence and also in the interview material we see certain ‘repertoires’. For instance, when we see discussions about smart cities, sometimes the act of measurement seems to be political and sometimes apolitical”.
“We tried to understand how these repertoires are structured, identifying five regimes: bureaucratic, scientific, political, engineering and civic.”
”Our idea is that these are separate possible ‘worlds’ basically, between which the actors can move.”
”For example, in each ‘world’ smart infrastructures work in a different way. From a bureaucratic point of view, such as in a public administration, the whole purpose of sensors is to provide actionable data. The accuracy of this data is not a great concern, and what can be learned from the environment using sensors is not very important. The interesting thing is to signify that this data can be operationalised.”
”On the other hand, for engineers designing devices, the accuracy of sensors is the central concern. So sensors are required to be accurate.”
”These regimes are separate imaginaries of ‘smartness’. Our observation at the moment is that there are some kinds of internal networks of ideas each with their own logic from which one can move from place to place, that not necessarily connected to each other.”
”The central purpose of CANDID is to connect these different perspectives.”
”It was an assumption at the beginning of the CANDID project that SSH scholars had different perspectives than ICT practitioners but we actually found that these groups learn from each other, and can reproduce each other’s discourse, although that doesn’t necessarily mean that puts them in dialogue.”
”We found ICT practitioners can talk a bit like social scientists about ‘smartness’ but just because they can understand and reproduce another type of discourse doesn’t necessarily change the way they do their engineering.”
Fleshing out these regimes
The impression gained when interviewees talked about how sensor infrastructures would work, such as in urban areas, was they would express different expectations for these technologies.
Maxigas found it interesting that these expectations seemed to correspond with ideas concerning expectations formulated by one of the founding sociologists, namely Max Weber.
“It seemed expectations for smart technologies that veered toward bureaucracy had some common elements.”
”As Max Weber would say, public administration has to answer to the requirements of efficiency, rationality, precision, reliability, but most importantly neutrality.”
“This was interesting to us because although ‘smart’ infrastructures may be an emerging issue for sociology, on the other hand, you could also say the study of bureaucracy is as old as sociology itself.”
“So, if you use one to question the other we can mobilise a whole sophisticated body of knowledge.”
Subtle issues of rationality and bias
The formulation of ‘smart regimes’ and bureaucracy then enabled the team to account for some contradictions in the empirical data.
According to Maxigas, “A majority of the interviewees were quite ready to say that technology is as political as anything else in our lives, and that prejudices and biases are part-and-parcel of technological development, and are routinely encoded in the technologies themselves. Yet, when we asked them to point out how these abstract ideas manifest in the technologies they are actually working on, usually they had a hard time demonstrating that assumption.”
“In fact, when they explained how their own technology functions, they told us not to worry because technologies in their own projects are reliable, efficient, and they favour talking about rationality and any political bias is, somehow, neutralised”.
“Bureaucracy has been one of the cornerstones of maintaining the functioning of liberal democratic societies, such as in the separation of the legislative and judiciary from the executive branches of governments. The executive is supposed to be politically neutral but part of the democratic function is to respond”.
“So, if a politician is seen as not having values or doesn’t respond to citizens then he or she is seen as being a bad politician”.
“But, on the other hand, if a bureaucrat is seen to be paying attention to the concerns of citizens, taking decisions in a personal capacity, or to be doing the job according to some personal values, this can be considered a problem for democratic society, and this bureaucrat can be easily seen as corrupt.”
“There are different things you have to do as a politician compared to a bureaucrat.”
“What we expect from politicians is a kind of commitment to values, but bureaucracy is based on technical expertise.”
“It’s always been a question of how a universal idea of democracy and technical expertise, such as bureaucracy, can co-exist in a social system”, a tension we’re currently seeing in the United States for example.
“These actors are facing new questions in thinking about sensor infrastructures as bureaucratic structures face being replaced by autonomous machines. So, in the case of the sensors being used to optimise the routes of trash collections, the city would have a bureaucratic office organising the labour. The output of the sensing infrastructures would replace the work of this office, but the question is, whether it is replaced with decisions that work in a different way or can this black box function according to the same principles as the thing it replaces.”
“Maybe if the thing it replaces is one guy who talks a lot with trash collectors and who organises the routes, it’s not clear if the computerised system does actually function according to the same principles, or fulfils the same expectations.”
Social science and humanities as party poopers
One of the broader ideas the Module 3 team is looking to raise at the Barcelona meeting is an analogy made of Social Science and Humanities scholars as ‘party poopers’.
“From the interviews with social scientists we’ve heard when it comes to intercourses around innovation they’re often relegated to the role of ‘party pooper’.
”The social scientist is seen as a barrier for the innovator, as purely a negative.”
“This is a frustrating situation, and one of the obstacles to interdisciplinary work across domains that we want to ‘debug’ via the CANDID project.”
“If this frustrating situation persists then whatever the social scientists say is going against some of the fundamental myths behind modernity, such as the idea that technological progress goes hand-in-hand with social progress, or the idea that there is a linear history of technology, and there are no alternative pathways, only an ‘arrow of progress’.”
Matching the presentation styles for product development
Another aspect the researchers compared was the manner of the respective presentations of SSH and ICT discussions concerning sensor infrastructures.
”Both present their questions and expertise as being very complex, but, at the same time, when dealing with external actors, engineers tend to come to the table with a claim that they bring simple and efficient solutions to complicated problems, while social scientists sometimes like to present themselves as being people that complicate things, offering a more ‘sophisticated’ understanding of issues.”
“This idea of complicating things does not fit neatly into product development, for example, because that’s about closing issues and finding ways to make overlapping interpretations of technologies.”
“In science and technology studies we speak about different ‘technological frames’ attributed to actors around a prospective technology, looking at different closing mechanisms used to stabilise a technology before it comes together as an identifiable technology.”
“Social scientists in this role are seen as a bit of a nuisance by some other actors.”
Framing social science as an enabler of innovation
The team hope to identify or create another positioning, to reboot this dialogue.
“Maybe social scientists can enable engineers to think about something that is really new.”
“Real innovation is not to just about making some more efficient machine but brings in aspects that had not previously been considered.”
“If we can frame such a contribution, so that it becomes an enabler of innovation, so social scientists can be seen as not there to stop innovation but to open up innovative pathways for technology development, we could perhaps consider sensor infrastructures in an even more disruptive way.”
“There is also a question of whether we can or should go beyond ‘modernity’, or develop a very different relation to technology. Despite such claims, however, what we tend to see when studying sensory infrastructures is that what happens is actors incorporate these systems into existing structures and mobilise them for the same purposes and in the same ways they’ve thought of previously.”
Consortia Meeting coming up in Barcelona
The 26 September consortium meeting will be an opportunity for the researchers in all three modules to consider these early findings and considerations.