#Human: How A.I. Chatbots Are Relieving Social Isolation

What does a good future city look like?

That’s the question being asked by OrganiCity, an EU-wide organisation which has funded sixteen teams across three cities (London, Aarhus and Santander) to explore answers. Organicity’s projects are driven by digital technology, experimentation and co-creation with citizens.

Colour-in City is just one London-based social enterprise that has risen to the challenge. Working out of Impact Hubs throughout London, the group’s name comes from ‘colouring-in the outlines’: a process of design and service development which has collaboration firmly at its heart. More specifically, they’re small team of service designers, data experts and public sector providers passionate about improving social wellbeing. I grabbed a coffee with Eunji Kang, Colour-in City’s design researcher and information designer to ask her more about their response.

Tech and Wellbeing

“Colour-in City’s focus is on wellbeing and how we can use technology to gather people’s perceptions and stories,” she tells me. “We are small but we’re asking big questions. And we’re particularly interested in the combination of quantitive and qualitative data.” When the team partnered with Lambeth Council (Lambeth is one of London’s most densely populated and socially challenged boroughs) and LEAP (Lambeth Early Action Partnership) they identified that Lambeth parents living in social housing were particularly affected by a wide range of social issues. They had identified their end users and collaborators.

The team’s challenge was to create and prototype a data collection service in just four months — the period of funding. But they were clear that the key to the project’s success would be fully dependent on the collaboration of Lambeth parents themselves. “In the beginning, councils were talking about practical support to improve physical spaces, such as providing better storage options,” Eunji says. “These practical solutions are an important part of housing services, but our research showed that parents also wanted a better understanding of what was going on with their emotions and wellbeing.” So how could Colour-in City sensitively collect data on Lambeth parents and use it to improve their wellbeing?

Time for a chat?

Isolation, identity and shame

Towards the end of 2016, they began a design process which involved sensitively scoping issues affecting the community. The team spent time in parents’ homes “to gain insight into their everyday lives.” What they discovered was a spectrum of issues which included a lack of space, (particularly in flats with children and teenagers), feelings of shame about living circumstances and social isolation. This was especially acute for parents without strong family or community support networks. Many mums who devoted their lives to their families also confessed to issues around their sense of identity and self worth. Many felt invisible to local services.

With these learnings, Colour-in City held two co-design workshops; one with parents and one with local service providers. Their aim was to understand, in the parents’ own words “what supports and what gets in the way of their wellbeing.” The team shared one particular mother’s story as it represented a wide range of issues they’d encountered, and whilst it was at the extreme the extreme end of overcrowding, (a single parent sharing one bedroom with her three children), “through her story we found experiences and emotions that other parents understood.”

Summary passed on to parents

‘Reciprocity’ and ‘Respect’

Two core principles guided this process: reciprocity and respect. After all, the project involved the collection and analysis of data — something that could easily feel cold and invasive. Through reciprocity the team understood that for those interacting with the service the experience should be friendly and warm; and respect should be shown for their stories, experiences and the value of the data being shared. From this starting point, the team could begin “designing a digital tool for parents to track, use and share data about their everyday experiences.”

A chatbot made sense in terms addressing three requirements identified in the parents’ workshops. Accessibility: they wanted something that could fit around their busy schedules and needs. Approachability: they wanted an interface with character, charm and empathy — something they could relate to. Finally, utility. They wanted something useful. If they were going to give up insights about their lives and experiences, they wanted practical help in return.

Empathy

Using the qualitative data collected at the workshops, the Colour-in City designers began to develop a persona for the bot with characteristics the parents approved of. They worked with a data scientist on its A.I. elements, developing the bot’s ability to recognise language patterns and respond in a conversational tone. Not only could the it ask parents how they were feeling, it would follow-up with questions and comments depending on the answer.

Visual data supplied to parents

In fact, when an early iteration failed to ask adequate follow-up questions, (when users told it they were feeling sad or overwhelmed, for example), the team was able to adjust it in an agile fashion. “I’m sorry to hear that,” or similar comments were added to enrich the interactions. Parents had also said they felt it was important that they could request support from a real human being if they needed to. Functionality was added so that by typing in #human into their thread, parents were put in touch with a real person offering support or advice. “Parents wanted to feel listened to and they wanted to have their feelings taken into account,” says Eunji.

A Friendly Face

In terms of look, the chatbot needed a friendly face. The designers took qualities gathered in the workshops and began to develop a persona and Squeezy was born. A cross between an ink-blot and a stress ball, Squeezy was cartoonish and playful — someone you could take your stress out on or open up to. He also looked young, which was important. “We encouraged parents to see our chatbot as a ‘baby’, ‘a young bot’, still learning, growing, needing our support, patience and input.” This meant that not only could Squeezy account for his glitches and limitations, he had a simulacrum of human vulnerability.

As users tested the prototype through various iterations, it became clear that Squeezy was collecting data which could give local authorities powerful insights into citizens’ lives. Broadly speaking, the were three ways Squeezy could do this. First, he was able to map qualitative data from users and find out what was concerning them. Second, he could offer empathetic or encouraging comments and point the parents in the direction of relevant resources. Third, after a couple of weeks, Squeezy could share with parents insights into their own interactions. And using machine learning and language recognition technology, he could interpret the parents’ data for them.

Parents testing the service

Feedback

Eujni confides that she was surprised at how much users enjoyed interacting with Squeezy. “I didn’t expect it,” she tells me. What was intended to be a friendly UX (user experience) actually ended up becoming a feature that some parents relied on. This is evident in the feedback — some of which can be seen in the short film below. “I’ve paid more attention to myself. It’s basically a brief diary of my emotions, my feelings, my thoughts,” says one mother. “I got tips from other parents so I made connections,” says another. The users were also aware of the positive benefits their data could offer local services. “Lambeth will know what parents need and what things should be in focus.” A comment that the bot offered “new opportunities” was representative of many users’ experiences.

There are ethical questions implicit in all of this of course. To get the Squeezy up and running quickly the designers had to integrate him with Facebook Messenger. Whilst this fulfilled the parents’ accessibility requirements, it meant that, as with so many apps, Squeezy’s creators were bound by the social media platform’s privacy terms. With data protection on social media such a sensitive topic these days, Colour-in City were clear with parents what the implications were for their data. The fact is, however, that as of September last year, there are over 30,000 chatbots on Facebook — Squeezy is far from alone.

And luckily his story continues. At the time of writing this, Colour-in City and LEAP have just heard they’ve secured funding to develop their A.I. bot on its own channel so that it will be fully secure and data protected. Squeezy will also get a bit of a makeover.

The Rise of the Bots

There’s surely a paradox at the heart of a robot designed to alleviate social isolation. Many would question what sort of a future we’re heading into, where artificial intelligence replaces human society to relieve our loneliness, depression or anxiety. However research shows that humans are quite comfortable interacting with chatbots as long as they know it is a bot. It’s when the boundaries are blurred that we feel unsettled. With the boundaries between machine and human intelligence becoming ever more integrated however, this will surely become a point of future tension.

Squeezy was only over intended to be a tool connecting parents to resources, services or one another. Colour-in City are still evaluating their results, but early signs suggest that, on those terms, Squeezy has been a success. Whilst technology develops at an inconceivable pace, these are still early days for AI and machine-learning bots. “A safe space for exploration” is how Eujni describes Colour-in City’s project. Meanwhile, inventor, futurist and Google engineer Ray Kurzweil, predicts we’re not far off seeing a chatbot pass the Turing Test. “If you think you can have a meaningful conversation with a human, you’ll be able to have a meaningful conversation with an AI in 2029. But you’ll be able to have interesting conversations before that.”