In my current role, I work at the intersection of smart cities and community engagement. There are two facets to this work. The first is defining the need to engage people in a smart cities context. In smart cities dialog, it is increasingly apparent that the role of the individual has become amplified as a counterbalance to the accepted emphasis on technology. After all, at the end of the day, cities are for people, not technology; the goal of a smart city should be to better meet the needs of those people, which can be difficult to achieve if we aren’t tapped into their desires, aspirations and beliefs.
The second facet, which I focus on in my work, are the tools and how we can use smart cities technologies as an effective way to improve existing community engagement practices, and in particular enable continuous sensing of community input as opposed to one-off engagement exercises. For so long, we have been bound by the box of community workshops, town halls, and written comment – paper and pen tools that are time intensive to participate in, lack excitement, and are often unrepresentative of the broader population. We have also been bound, although this is a separate conversation, by poorly devised engagement activities that lack specificity in purpose and often result in input that does not meaningfully add value to the project at hand.
More traditional engagement practices are meaningful for many reasons, such as creating healthy dialog, building ownership and buy-in. However, we should look more seriously at how we augment them with other ways of deriving information about our constituents. This can help us support better, more responsive decision-making, and ensure that community input doesn’t become an afterthought once a project has been developed, but is embedded in the planning process every step of the way.
As such, we would be remiss to ignore some of the tools and analytic abilities that have been cropping up to enable smarter, better, and more efficient cities. In particular, it would be worthwhile to explore whether they can be applied not only to city management, but also community engagement. To this end, I have been exploring a few tools and methods, laid out below:
- Mobile phone sensing. One of the key dilemmas of smart cities is how to harness and use big datasets to better understand people’s movements within a city. With mobile phone sensing, one can obtain location data collected by people’s mobile phones as they move around a city, which can then be further analysed to provide deep insights into how people traverse and use a city. Mobile phone data is often obtained through agreements with cell phone companies to provide anonymized data. Increasingly, data can also be obtained by apps that run GPS data in the background, or in some instances, residents may be willing to install specific apps that allow researchers to collect their data directly.Of course, this isn’t limited to mobile phones. Other data traces, such as credit card transactions, data provided by companies like Uber and Waze, and smart transit cards that enable us to track where people tap in and out of metro stations also provide a fine-grained data set that can help us better understand the needs and behavior patterns of local residents.
Potential community engagement applications: Mobile phone sensing and other methods of tracking human movements can be used to replace, verify or augment data derived from answers to traditional workshop questions such as “How do you think we should prioritise pedestrian improvements?” Mobile phone sensing may even replace self-report data obtained via traditional survey (or other) methods. By knowing how people travel, and where they go, we can more easily surmise how they may value improvements intended to facilitate mobility. If pedestrians are avoiding particular routes even though they provide the most direct path from their origin to their destination, that tells us something about how they perceive the walking environment of that route. Or perhaps we could focus on pedestrian infrastructure in locations where people are driving to destinations that are within walking distance, or where there is a critical mass of pedestrians on narrow footpaths. Instead of asking people “What would you like to see developed here?” we could have a prior understanding of which amenities local residents are traveling outside of their neighbourhood for (e.g., parks, nightlife, grocery shopping, coffee) to assess not only what people say they want, but how they actually vote with their feet.
- Participatory sensing. A cousin of more passive mobile phone sensing, participatory sensing requires people to actively participate by collecting and submitting data, thus enabling “crowdsourcing” local knowledge. Oftentimes, participatory sensing is conducted via mobile apps, whereby people use their phones as a sensor to generate data, such as noise levels, photos or personal observations, that they then submit. Sensors that are embedded within most modern mobile phones include microphones that can serve as audio sensors, video and photo, GPS location information, pedometers, accelerometers that indicate the speed with which one is moving, and a myriad of other sensors that can be paired via Blootooth or wired connections, such as air pollution or biometric sensors (Kanhere 2013).There is also a separate branch termed “citizen science,” oftentimes used in environmental settings, whereby engaged residents will set up their own network of sensors, such as air quality or water quality sensors, as a method of providing on-the-ground conditions to policy makers and regulators.
Potential community engagement applications: Participatory crowdsensing could be used as a mode of collecting data about a neighborhood, as opposed to asking residents for similar information in a community workshop. For instance, instead of asking people to recall where they don’t feel safe, or places that are dirty, people could provide data on areas where they don’t feel safe, or areas where they see litter, as a method of identifying locations that require improvement.
Also, there are broader applications as well to inform broader city planning processes. For instance, as cities become more concerned about noise pollution, they can rely on residents to help them document noise levels across the city to create a comprehensive map of decibel volumes. Similar work has been done with potholes, whereby residents in Boston can download the StreetBump app that can sense motion when a driver runs over a pothole, and send that real-time information to the city for repair and to inform longer-term infrastructure planning activities.
- Sentiment analysis. Sentiment analysis is an interesting method that mines opinions expressed on social media platforms, such as Twitter or Facebook, and analyses those opinions to provide a general gauge of how people feel about certain topics. One of the strengths of sentiment analysis is that it is relatively unfiltered, and thus may lack the bias that may present when people know that their input is being used for a specific purpose. Sentiment analysis most often makes use of machine learning, which can enables computer programs to ‘read’ the content of the social media post and detect patterns and spikes in activity. Thus, researchers have been busy constructing sophisticated models that can quickly parse through social media posts, identify key words and then indicate whether the emotion (or increasingly, emoticon) being expressed about the key word is positive, negative or neutral, as well as the intensity of that emotion.
The use of sentiment analysis in the prediction of election outcomes has been particularly well publicized. For instance, a 2010 study by the TU Munchen analysed more than 100,000 tweets containing a reference to political parties and politicians during the 2009 national elections and found very close correlations between the number of tweets about a political party, political polls and the election results (Tumasjan et al 2010). On the simpler end, Kavanaugh et. al (2012) conducted a simple content analysis of “likes” and posts to the Arlington County government Facebook page during a two month period, and were able to use this data to draw inferences on the topics that were of most interest to its online constituents. They concluded that social media can be an effective method to detect meaningful patterns and trends of “issues of concern for public safety or general quality of life (e.g., traffic, air quality).” (Kavanaugh et al 2012).That being said, there remain problematics with sentiment analysis using social media. Social media users are not representative of the overall population; results are highlight sensitive to the methodology used to parse the social media posts; and researchers haven’t yet cracked the nut on how to accurately translate social media posts into sentiments, or to differentiate between “real” posts and “fake” posts. Keeping these limitations in mind, there are many potential uses of sentiment analysis as a way to help identify public opinion.
Potential community engagement applications: The first application that comes to mind are long-term vision plans that many governments undertake (see for instance Greater Helsinki Vision 2050, Imagine Boston 2030, 2030 Seoul Plan, etc.), which are often based on understanding the issues that are most important to their constituents. Sentiment analysis could also be a method to derive general feedback about city performance in key areas. Currently, many cities employ large scale polls and surveys, often by telephone, to rank government performance on a number of indicators or to understand opinions about controversial topics. These data could be supplemented with data obtained via big data solutions such as sentiment analysis.
- Emotional Cartography. Emotional cartography is an approach to understanding and quantifying the human body’s physiological response to certain emotions, such as excitement, happiness, fear or sadness. This response can manifest in many different ways, such as activation of different regions of the brain, amplified heart rate or increase in sweating, and can be measured by one of the ubiquitous tools in smart city speak: sensors. Unlike mobile phone sensing, which provides data on where an individual is located geographically, emotional cartography is based on how the individual feels when they are at that location. By having people wear tools such as wrist bands, galvanic skin response sensors or more elaborate helmets that are wired to measure brain activity (EEG), researchers can obtain detailed data on stress, fear, anxiety and other emotions from people as they traverse a city. Again, by relying on a human body’s response to the environment, emotional cartography can generate real-time, natural feedback on where people feel safe, unsafe, happy, unhappy – key inputs for broader planning activities.
However, emotion sensing still remains largely in the research realm. While there are many on-going research projects in this field, the products are not off-the-shelf solutions that can be deployed and would likely be best done in partnership with an academic or research institute given the complexities in study design and sensor calibration.Potential community engagement applications: The use of emotional cartography could supplement questions that are often asked of participants in community workshops such as “Where do you feel safe?” or “What places do you like?” By developing a comprehensive map showing point-by-point people’s reactions to places, we can have more in-depth feedback to guide urban policy and planning. For instance, one could generate a map of the places within a city where bicyclists feel the most stressed, study the factors that may be contributing to the stress and prioritize those areas for improvements.
- Chatbots. Chatbots are computer programs designed to conduct meaningful conversations, using key words and cues to engage with and respond to people. Chatbots can be deployed online, through chat interfaces such as Facebook Messenger or even through simple text messaging. They are increasingly used by companies and institutions for a wide variety of purposes, ranging from answering frequently asked questions to helping customers reorder their favorite items online. The simplest version of a chatbot is one that is hard programmed to respond to certain phrases or words with corresponding phrases or words (e.g., if you type “hello,” the chatbot will know to respond with “hello, how are you?”). More complex chatbots use artificial intelligence and machine learning to enable them to have more varied conversations, with the ultimate goal of passing the Turing test – meaning, one cannot distinguish if one is talking with a human or a computer.There are many platforms available, often at a nominal cost or for free, that enable one to program their own chatbot. Once deployed, chatbots can be available 24/7 to engage with people from the comfort of their home, thus lowering the barriers to data collection.
Cities have been experimenting for a while with ways to engage with residents online and enable them to reach residents who do not have the time or desire to attend public meetings. For instance, there are several online platforms where cities can post questions for residents to respond to (see for instance MindMixer) or even enable residents to provide feedback on an online map, by placing pins at specific locations and writing comments such as “the surroundings here make pedestrians miserable” (actual feedback from the Helsinki Vision 2050 planning process).
However, these platforms have struggled to gain popular usage, as they still require one to have a computer and internet connection, go to a website, register and complete a set of prescribed activities.The benefit of a chatbot is that it often communicates via existing apps on a mobile phone, or even text messaging, thus enabling people to communicate without an internet connection or having to download new apps. Deployed in an urban planning or community engagement context, one can imagine how chatbots could be used to emulate the town hall or community workshop setting, enabling people to provide input from the comfort of their phone or home through engaging in a virtual conversation led by a chatbot. Further, the conversational tone of a chatbot can be effective in encouraging people to provide meaningful and honest responses.
Potential community engagement applications: Chatbots could be used to enable people to provide feedback on plans, projects or even services by being programmed to ask the participant a series of questions, such as “What do you like about your neighbourhood?” or “Do you think bicycle lanes would be a good idea on 4th Street? Why or why not?” Some cities have also deployed tools similar to chatbots to solicit feedback “in the field,” for instance by placing signs in public spaces asking people to text in their ideas for improvement, thus enabling a 24/7 feedback loop with minimal barriers to entry. However, these tools remain largely one-way inputs rather than a two-way dialog. The next frontier will be to consider how to create chatbots that can process the inputs, and then follow-up with additional questions to obtain more detail and context to position their responses.
These are the tools currently at the top of my mind that could be a potential path to augment current engagement practices, focused not only on a single point of engagement, but rather continuous engagement that allows citizens to input their “voice” into city making and city planning processes on a day-to-day, real time basis. To date, most of these applications have taken place in a research setting – thus, the next question is how do we bridge this research with practice and bring some of these technologies to bear to create smarter, more user-informed cities?
– Julienne Chen, Sr Research Associate
Kanhere S.S. (2013) Participatory Sensing: Crowdsourcing Data from Mobile Smartphones in Urban Spaces. In: Hota C., Srimani P.K. (eds) Distributed Computing and Internet Technology. ICDCIT 2013. Lecture Notes in Computer Science, vol 7753. Springer, Berlin, Heidelberg
Kavanaugh et al. (2012) Social media use by government: From the routine to the critical. Government Information Quarterly 29: 480–491
Tumasjan et al. (2010) Predicting Elections with Twitter: What 140 Characters Reveal about Political Sentiment. Proceedings of the Fourth International AAAI Conference on Weblogs and Social Media.