Monthly Update 2021-04 and 05
This edition contains: two blog posts, some reflections on our recent product experiments, and an update about our recent workshop.
Bramble Blog Posts
This update, we have two blog posts for you: Feelings are what we have in common, and “I care about…” - and on a brand-new website, as well! (Check it out!)
Bramble Blog: Feelings are what we have in common
In the flurry of dealing with your own feelings and trying to manage your children’s behaviour, it’s easy for parents to forget that kids have feelings too. No matter how old we are, our feelings are what we have in common with one another.
Check out our blog post where we debunk 3 common misconceptions that parents have when it comes to their kids’ feelings.
Bramble Blog: "I care about..."
Fill in the blank - “I care about…”. When it comes to your child’s behaviour, what’s the BIIIG picture behind what you’re asking them to do?
What’s a BIIIG picture, you ask? Well, the ‘little picture’ refers to what you want your child to do. (You can probably think of a list of these things already!) The BIIIG picture refers to why you want them to do it! Read on to learn more about the BIIIG picture, and how to share it with your child.
Product Experiments
Cycle 4
In our last update, we talked about our goals to help parents and children reach an actionable solution through their conversation, and to implement APIs which would help us integrate further interventions in the conversation. After testing these out in Cycle 4, we came to the conclusion that this was a strong step in the right direction, given the higher and more consistently high scores from our post-test user surveys!
Insights from the Cycle 4 Experiment
From our first foray into helping parents reach an actionable solution, we identified some concepts which were very helpful for parents (which we’re excited to share about in future blog posts), as well as some common pitfalls within the problem-solving process.
We identified 2 key stumbling points within our conversational flow which prevented parents from reaching the value we wanted to provide:
Not being able to direct the parent’s attention to when the child said something key, with the parent either missing or dismissing what their child said; and
‘Just Do It’ solutions, a kind of solution which - as its name implies - ends with the parent saying ‘just do it’, e.g. ‘just do your homework’, or ‘just be more disciplined’.
The preliminary tests as to whether we should go further with API implementation also produced good results - we determined that we would be able to gather enough user data while users were having the conversation to suggest preliminary interventions (for instance, using facial sentiment analysis to redirect the conversation if they looked tense or angry).
Overall, we found the Cycle 4 results very energising, especially after receiving one conversation recording from a parent-child pair who seemed to reach a breakthrough in their relationship via Bramble! The parent-child pair has thankfully allowed us to share their recording, which is accessible here: https://www.letsbramble.com/video-tour
Cycle 5
With our energy running high, we hoped to further refine the conversational flow to meet the identified needs of parents and children. We hoped to use specific selectables for challenges that the child faces to understand the precise shape of the problem parents and kids were talking about, and adjust the flow accordingly. At the same time, we hoped to enhance user experience by smoothing out grammar and the automatic recording process.
We also took on a new challenge this cycle - live tests, where we would watch a parent-child pair having a conversation in real-time over zoom, and intervene (by changing the content of the app) based on where we thought the conversation would bring the most value. The goal was to understand how users would respond to our conversation facilitation tool in the ideal situation where the app could do everything a human could, if a human were mediating the conversation live.
Insights from the Cycle 5 Experiment
In Cycle 5, we managed to refine many of the things we started in Cycle 4, pushing personalisation to a new level. Our efforts to focus on parent profile personalisation and problem definition personalisation were particularly rewarding in this cycle, with fewer parents missing or dismissing the key element of their child’s struggle.
However, we realised we failed to make much progress in terms of helping parents reach an actionable solution, with many parents continuing to adopt ‘Just Do It’ solutions. We determined that this was a big area we wanted to double down on in the next cycle.
Insights from the Cycle 5 live tests
We found that it was harder than we initially thought to intervene effectively using our live test system. Instead of being able to just speak to pause or redirect the conversation, we had to use specific commands to change specific nodes in the conversation graph, and type all of it out, while the conversation keeps going on - so there was quite the learning curve!
Still, there were a few instances where we felt the live interventions improved the conversation, for example, adding more specificity to certain prompts based on what they were saying, or short-cutting the conversation if it was feeling draggy. Overall, while the live tests were slightly more personalised, we didn’t feel that we managed to achieve the ideal case of the app being able to do everything a human could.
Cycle 6
Ultimately, Cycle 5 left us firm in our belief that personalisation and specificity in prompts throughout the conversation would allow for a much smoother user experience. We hoped to get there by improving parent profiles, and tying different elements of the conversational flow tightly in with each other with visual cues and additional signposting.
As such, our goals for Cycle 6 were to:
Improve the parent profiles by allowing for more in depth questions through conditional logic in the parent questionnaire.
Introduce some useful concepts (e.g. “You” vs “I” statements)
Strengthen the link between boundaries, wiggle room and problem solving
Help users create more concrete, actionable solutions
Highlight judgemental/loaded adverbs in the text input fields, and explain what the effect of using such words might be.
Use ML to classify statements users said as critical/judgemental, and develop paths of the graphs that users would be redirected to if many of these statements were detected
Analyze users speech and facial expressions to determine their emotional state during the conversation.
Add some functionality to the live conversation system
Insights from the Cycle 6 Experiment
Overall, we didn’t feel that this cycle brought us significantly closer to our vision for the product. However, when processing the test data for this cycle, we managed to bring many of our previous takeaways and ideas to maturity, to be developed in Cycle 7.
One notable example of this was: after a few cycles of middling results, we determined that the part of the conversation where parents shared their big picture didn’t seem to provide enough catharsis for the parent, and also failed to safeguard against the parent placing all their fears on their child. As such, this was a key area which we wanted to innovate on in Cycle 7.
Another area we managed to bring to maturity was the issue of reliably bringing parents to what we called the ‘first level of value’, by helping the parent reframe the child’s behaviour to be the result of ‘skill’, not ‘will’, and peeling back the curtain to reveal why the child found performing the desired behaviour difficult. But we found that the conversation flow was still inconsistent in its ability to deliver the next level of value - a thorough exploration of that in order to find all the clues and insights they need to come up with an effective, actionable solution. As such, we thought it would be key to identify all possible ‘next levels of value’, possibly tied to parent profiles, and to map out paths to get different parent-child pairs there.
Although the added tightness and personalisation of the problem-solving process to the precise ‘hardest part’ identified previously did help some parents and children reach a more actionable solution than before, we still felt dissatisfied with the actionable solutions that parents and children were coming up with. Upon reflection, we realised that this was because we noticed a tendency for one party to suggest an idea and for the other to just “okay” it rather than really engaging in a true back-and-forth, evaluative brainstorming process. This is very much a cultural norm in Singapore, where parents tend to rely a lot on their parental authority, but we wanted to bring parties closer to a place where they would bounce ideas tightly off each other, really listen to each other, and respond to each other’s points.
While thinking about this, we also crystallised our thoughts on how to avoid ‘Just Do It’ solutions. We determined that a significant factor behind parents’ belief that they should ask their child to ‘just do it’, and that a solution where their child didn’t ‘just do it’ was a failed one, was because there was insufficient co-creation / mutual involvement in the solution - and parents didn’t know how to get involved, or how they wanted their kids to get involved.
Insights from the Cycle 6 live tests
We didn’t see that much improvement in the level of impact we were able to create for our live test users, given our current tools. As such, to ‘break free’ from the limits of the current tools, we plan to have Bramble Guides unmute themselves during the live test conversations and directly intervene. We hope to observe how parents and kids respond to that highest level of responsiveness, personalisation and intervention.
Cycle 7
Given all our insights from Cycle 6, our goals for Cycle 7 were to:
Prepare the app for release on the app store
Improve the way the app corrects grammar using an AI API instead of only using our homebrewed system
Improve the accuracy of speech to text as it is the foundation of our ability to understand the content of the conversation and intervene on that level.
Optimise the entire conversation for 1 topic (learning/homework/revision/exams)
Deliver the next level of value
Increase levels of co-creation and co-involvement in the solution
We’re currently in the process of going through test results, and are eager to let you know how they went!
Workshop and YMWB Collaboration
On top of our usual development work, Bramble partnered with the Youth Mental Well-Being (YMWB) Network during Cycle 7 to create a workshop entitled Preparing for PSLE: Understanding and Managing Your Child’s Stress. In doing so, we hoped to establish a friendly relationship with YMWB to open the door for future collaboration, as well as to reach a broader group of parents and start gathering ‘wild’ users.
In this workshop, we presented a new acronym to help parents talk to their children: INVITE. Check out an infographic we made on the acronym here:
That’s all for this update! Thanks for supporting Bramble. Stay safe, and see you soon!