‘Hey there! Is this a good time for a chat?’
From reciprocity to respect; how our friendly chatbot is generating rich data, community connections and unexpected wellbeing outcomes.
‘Hello, hey, hi! My name’s Squeezy because I’m like a stress ball you can squeeeeeeeeze to get out some of your stress, worry or anxiety. I’m still a young ‘bot so please be patient with me. The best way I learn is if you tell my makers about my mistakes. Looking forward to chatting soon!’
We’re Colour-in City, a team of service designers, data analysts, public sector providers and citizens exploring how to create a good future city using digital technology, data, co-creation and experimentation.
Our ambition is to co-design a digital tool that helps reduce stress and anxiety and improves wellbeing for parents in overcrowded social housing.
The quote and picture above is from Squeezy’s profile. This introduces the parents we’re co-designing with to an ‘internet robot’ ready to help them track, use and share data about their everyday lives with the local services that support them.
Our hypothesis is that if parents and services better understand what supports and what gets in the way of wellbeing on a localised, individual level, they’ll be in a better position to identify, shape and access the support they need.
In this blog post we’ll explore the process of developing Squeezy together (beta version), focusing in on how we’ve built collaboratively for reciprocity.
Reciprocity through chat
‘Reciprocity: The practice of exchanging things with others for mutual benefit’
‘…from Latin reciprocus, moving backwards and forwards’
The idea of creating and using an AI chatbot to collect data was the most popular of various digital options we explored with parents. It quickly ticked their usability needs, which included:
- be easy to use
- feel fun, not another survey or questionnaire
- work on their smartphone or desktop computer
- be reciprocal, give something back.
The need to embed reciprocity was one of the most important design principles to come from parents. A chatbot offered huge potential to deliver this. It allowed for back and forth conversation; a 1:1 ongoing relationship; a means to collect data and provide data in response.
By systematically building up our understanding of parents’ needs through a human-centred design process, we’ve been able to see how reciprocity can and must run through the whole of their relationship with the chatbot. This means we’ve designed for:
- Reciprocity of data: parents share personal information, they get clear, useful analysis back
- Reciprocity in kind: parents give data that’s useful to local services, they want useful data back, eg about kids’ activities, answers to their questions
- Reciprocity in conversation: parents give data, the chatbot responds appropriately
- Reciprocal relationships: new ways of connecting are set up that create mutual benefit, between parents and parents, parents and services.
Building in reciprocity together
We developed the first iteration of our chatbot as a design team based on research into wellbeing frameworks, insights from parents’ experience, and their feedback on inspiring data tracking tools and visual reporting options.
Creating a chatbot gives us the opportunity to experiment with different types of data, using quantitative and qualitative research and analysis. This includes free text, emojis, keywords and images, as well as traditional scales.
With the initial prototype ready, we asked parents to help us improve it. Here are some of the ways their feedback is embedding that sense of reciprocity.
Make the bot kind enough to share happy and sad moments with
Parents wanted the chatbot to be fun, informal, to ‘be a robot’ rather than pretending to be human. We used this in the visual identity and voice, aiming for cute, friendly and plain English. We were also interested in how to bring play into the normally serious world of data collection.
Parents’ feedback and what we’re testing as a result:
1. Squeezy’s ‘cheeky’ style had parents laughing. They wanted the more formal standardised questions made friendlier and easier to understand.
What we’re testing: We‘ve softened the language around these questions, tagged them all as #wellbeing, and added emojis.
2. Its responses weren’t empathetic. Initially it asked a wellbeing question, gave a generic thank you, then went to the next question. Some felt this rude, especially when they’d given a sad response.
What we’re testing: The bot can’t yet know if it’s been told a sad story, but we’ve reworded responses and included a reminder that the bot is still learning in order to make it feel more sympathetic.
3. It was too pushy too soon. Some were annoyed that the bot asks for an image to describe an answer, before they’ve developed rapport with it.
What we’re testing: We’ve rejigged the conversation flow so the bot asks about the image at the end of the conversation, and are clear this is optional.
Help parents connect
Parents were keen to connect on issues in common. A chatbot didn’t seem to enable this, so we created a private group that we could link to.
Parents’ feedback and what we’re testing as a result:
1. The group is hard to find and not connected to the chatbot.
What we’re testing: We’ve added links to this as part of the bot’s conversation flow, including asking if parents want the bot to post questions or tips for them.
2. There isn’t much parent interaction on the group, but where there is it helps. For one parent, a tip helped narrow her job search and feel connected.
What we’re testing: We need to build on this. Online communities benefit from facilitation. We have limited resource, but aim to do more in the next iteration.
3. Simply having Squeezy to chat with makes some parents feel less isolated and more supported.
What we’re testing: This was an unexpected and powerful finding. We want to help parents feel supported but be wary of creating dependence. Parents can type keywords HUMAN or HELP at any point in the chat for a human reply. We’ll test how this scales with a larger group. Could we use the bot to connect parents with in-person support? As an experiment, we need to design for the end of parents’ relationship with the bot in a supportive way.
‘I like it, it’s like my friend. It’s my getaway.’
Parents want to receive something back for the data they share. This includes signposting to support, advice and ideas as described above.
It also includes their own data, clearly sharing this back in a way that helps them to track their wellbeing.
We’re on our first iteration of providing these kinds of visual data reports. While we’re currently limited by a small data set, as parents input more data and we expand our group of testers we’ll be able to visualise more.
Here are some early examples. We’ll get parent feedback on this soon to share in our next blog.
From overwhelm to control?
While not directly related to reciprocity, another important design principle was to help parents feel more in control, less overwhelmed.
We’re already getting feedback that simply chatting with the bot may be supporting parents with their wellbeing. It will be interesting to find out whether the act of tracking and noticing their wellbeing may in itself have an impact on those feelings of control and overwhelm.
‘I’m enjoying my chats with Squeezy. Normally, I go through my day, and i just ‘Do’. Me being aware of what is on the forefront of my mind motivates me to want to sort it out.’
‘Squeezy doesn’t insist that you chat. But you don’t feel alone. I feel secure and I can ask a question. Sometimes you don’t have someone to speak to. It makes you feel relieved.’
Respect for people’s stories
The reciprocity we’re designing for and that parents have asked for is, in a sense, about respect. Respect for the stories, the experiences, and the value of the data they are sharing with us.
While respect for personal data is important to every researcher, often the results of data collection aren’t felt in a direct way by those sharing their experience.
In our small-scale experiment, working directly with parents and their local services, we’ve got a chance to do this differently. Building in reciprocity at every interaction with the chatbot is one way we hope to show our respect in a visible and timely way.
Our main challenge has been the limited time available to make and test this prototype (this is a four month funded experiment).
To get the chatbot set up quickly we’ve used a chatbot integration with a private messaging service. While this has definite advantages of ease of use and access, it also means we’re bound by the data privacy rules of both providers.
We’ve explained this clearly to parents, but for the final product, ideally we’d build a bespoke chatbot. This wouldn’t rely on external providers, so we’d have full control over the chat experience and the data generated.
But the limitations of running a short experiment are far outweighed by the benefits gained from being encouraged to ‘experiment’. This ethos opens up the making process to genuine collaboration and curiosity, to testing new things, to being patient, and to keeping a critical eye and a sense of humour.
‘To be honest with you, I really wasn’t expecting it get this far in such a short time. Squeezy has become something really productive and useful to me.’
By mid-February we’ll have more data to analyse, more developed visual reporting, and we’ll be learning from parents about what worked, what didn’t, and what they want to happen next.
We’ll share all this on our next blog, as well as discussing the wellbeing questions we asked and the data analysis methods we’ve used.
Blog post authors: Emma Field, Rebecca Birch, Denise Xifara