Researching, Designing and Supporting Applied Behavioural Analytics in Parenting with Voice User Interfaces

service design

ux research

HCI

Fondly

2021

Project brief

Fondly is a conversational agent mimicking human conversation in Applied Behavioural Analytics. It was designed to use Home Economies, Tracking Routines and Token Rewards Systems to assist parents in parental skills development for early childhood (2-5 years).

timeframe

6 months

industry

Parenting Technology

01 context

It is widely recognised that young children often showcase behaviour that might be of concern to adults. However, in recent years, researchers have examined the clinical significance of behaviour problems present in early childhood. They are turning their focus on how these problems can be considered as potential indicators of more long-term difficulties. This signals a shift towards early intervention and behaviour modification. One of the most successful techniques for behaviour modification are token economies. However, they are not without their own issues. This project aimed to explore how technological interventions can help in supporting Applied Behaviour Analytical techniques, namely, token economies and reward charts. Based on existing solutions, literature and psychological research, I had seeked to design an accessible, easy to use solution. Designing a conversational agent with elements of gamification in the behavioural interventions process could support family routines and parenting more cohesively.

02 background specification, research and analysis

My undergraduate thesis aimed to develop a tool to support applied behavioural analytics through the use of a platform like Voice-User Interface. The design thinking process was strictly followed throughout the course of this project, beginning with investigating what the users needed through literature and fora analysis. The findings of this stage, combined with a review of assistive technology in HCI, was used to inform the define stage of the project. The define stage of the project was crucial in understanding the requirements that the proposed solution would need and fulfil. Ideas were explored through mind-maps and an interactive prototype followed soon after. In the technologies analysed during the literature review, the solutions seem to fit in a direction that promotes monitoring technology to facilitate care for others. Trailblazers like MOBERO and WAKEY were used as a basis for introducing the core theme of this thesis: Monitoring and Supportive Behavioural Analytical Technology. For design purposes, Token Economies and Gamification were chosen to be built upon. Research conducted has shown that parents often struggle with setting up token economies and keeping track of all the different tokens. This had impeded the effect of behaviour change. Building on these findings and assuming that an electronic automatic token tracking system would be easier to maintain and use, this project aimed to examine possible solutions for providing appropriate tools to aid parents in creating, managing and redeeming tokens through the use of Information Technology.

Themes appearing throughout the review

Thematic Codes Sorted by Context

Codes Occuring

As part of the project, a literature and fora analysis was conducted to gather data on how token economies have been implemented in a parenting or a teaching context. This analysis was done to shed light on what problems parents are facing in implementing, setting up, and maintaining token economies. The literature analysis was conducted on research showcasing existing practices, general issues and failures of token economies. An additional review was conducted to find a suitable alternative to the issues that parents face while implementing ABA techniques. An example of this kind of technology is ClassDojo, used by educators to address social skills and target behaviour using a token economy. A research paper indicates that ClassDojo was highly valued by teachers because the technology allowed them to streamline their work process, and easily administer tokens without the need of physical tools.

03 defining requirements

Creating an immediacy of reinforcement is highly important in token economies as delays in administering rewards could diminish the effectiveness of the token economy. Another important discovery was the lack of routines in paper-based methods. The lack of routines and behaviour chaining techniques is not grounded in the theory of habit formation, as self-monitoring is only effective if regular reminders and backup notifications are implemented in case of routine changes or lack thereof. Following the design thinking process, the literature and fora review stage provided an understanding of what the users need. The requirements deigned most relevant were:-

04 mind map

To better visualise the problem statement and gain a deeper understanding of the proposed design solution, the mind mapping process was used. These mind maps served to inspire base-level design. The original problem statement served as the central point, which further branched out into a free form mind dump :

05 conversational architecture

In order to design the proposed solution after understanding the requirements and ideation, it was crucial to understand and design what the conversational architecture would look like. The conversational architecture was designed with the available tools in mind. For every interaction, the information is fed to an NLU component that extracts the intent(the user’s intentions, and the entities intent). An example interaction may be “How is the weather going to be tomorrow?” In this context, the user’s intent may be defined as the user’s intention to get information on the weather, while the entity modification is the specified time, in this case, “tomorrow.” The extracted intent is then fed into an API from an external service. For this project, Voiceflow, NLU training and Dialog Management AI were used as the API’s, while the data was stored in Google Sheets and Airtable to keep track of the tokens, rewards and reminder options available.

06 conversational flow

Voice interfaces are now implicitly expected to be conversational instead of being dependent on commands or one-turn exchanges. The VUI Heuristics proclaiming that content should be findable, usable, communicative, learnable, controllable, valuable, credible and delightful were followed to the letter. Creating a visual diagram of the conversational flow helped indicate the different milestones that the user would reach.

07 implementation - starting blocks

The prototype design implemented the primary requirements outlined in the defining requirements stage, and followed closely to the conversational flow depicted above, namely, the requirements to monitor and track a token economy, set-up a token economy through the use of a technological device, and have reminders and check-ups to ensure that the token economy was followed diligently.

07.1 implementation - create chart

The create chart flow for this prototype focuses on gathering as much as information as possible from the user and stores it on a Google Sheet spreadsheet for tracking purposes. The creation of this user flow closely followed the guidelines highlighted in the handouts by Kazdin in his 'Parent Management Tracking Handbook'. The user indicates their name, and a nickname for their child. The VUI prompts the user to set out 4 target behaviours to want to monitor, before asking what their expectations for earning tokens are. The users must input an expectation for each of the target behaviours they identified. This information is directly fed into the database, before the VUI prompts the user for permission on setting up reminders and daily check-ins. In essence, the create chart flow is used for extracting Target Behaviours, Establishing Expectations, and finally setting Reminders.

07.2 implementation - Manage chart

Similar to the create chart, this particular flow is populated with intents. This flow allows users to add tokens, redeem tokens, change rewards, change reminder preferences, check token status and finally change targeting behaviours and established expectations. By creating intents, this means that the VUI runs without much user prompts, acting more like a conversational agent rather than a hard-coded machine. By informing the users of the options, the VUI waits for the user utterance that will trigger the intent.

07.3 final conversational flow

The tool used for designing this prototype was VoiceFlow. The reason for choosing this particular tool was based on the understanding that there aren’t a large number of VUI design and prototyping tools available. The other alternative to using VoieFlow would have been developing an alexa skill from scratch, however, this was outside the scope of this project.

create chart flow

manage chart flow

to conclude

An important aspect of this project was to understand how Voice User Interfaces add, store and manage data. The majority of the project was focused on writing a research paper on how Gamification can help carry along applied behavioural analytics. A key part of the project was understanding how to facilitate and organise workshops with vulnerable children, and how to bring parents into the conversation, when this is something the family has already struggled with.