Botocracy.
Superbots, Mindfulness and Trust-Models.
Challenge: How do we create a bot system that not only reflects and supports the user’s intentions, but also unlocks previously unimaginable collaborative and expressive possibilities?
Outcome: Prototypes that explore a new framework for bot design, grounded in designing to support user intention.
Group Members (alphabetical order): Damian Gill, Nicholas Jayanty & Christine Meinders
OUR APPROACH
Our primary goal was to use bots and superbots to design for our intentions. We framed our research approach by using the Time Well Spent project developed by Tristan Harris and James Williams. A secondary goal was to explore playful design opportunities by mixing and matching apps, bots, and other media in larger bots, or containers of intention.
RESEARCH
To explore this idea, we engaged in a multi-bot user study:
Mission: To more closely match the user’s intent.
Premise: The multi-bot concept empowers users to tailor their interactions with technological systems more effectively than monolithic AI.
Possibility: multi-bot interaction as a tool for creative/expressive engagement in technological systems.
In our research, we wanted to:
gain insights into our users’ intentions
explore how users might express their intentions in a multi-bot environment
discover the expressive/generative possibilities of a multi-bot universe
To achieve this we engaged in field-study interviews and a generative bot activity. Our questions focused on past behavior and a current focus on intent and productivity. We also asked users to design multi-bot environments.
For the bot activity, users were given evocative “bot-cards” to compose their super-bot team. They are also provided blanks to fill in their own concepts. Once the user composed the super-bot, they were prompted to devise scenarios where the selected bots interact with one another. We tested these cards in Los Angeles, CA with individuals, and groups.
We used the output from the card bot set to create personas and to inform our video prototype.
DESIGN OPPORTUNITIES
Based on our research, we have identified further opportunities in this design space:
To create a system for multiple intentions and bots
To create an interface which incorporates images of these bots with intensity level sliders and mute options
To create privacy settings which allow only certain bots to share select information with other systems/users
WHAT WE MADE
PROTOTYPES
In these three video prototypes we see how a life of intention plays out for Mia and her balancing of work, exercise, relationships and play.
In video 1, Mia modifies her standard template bot space. She is able to design and train her bots to reflect her intentions and purpose.
In video 2, Dev's bot allows Mia's taskmaster bot access to the schedule and interest data for Dev in order to schedule a weekend getaway for themselves. Once their data is combined, Mia's taskmaster bot (a superbot) delegates to the rest of Mia's bots in her botspace, to find a location and activities for the weekend. The outcome of the first suggestion did not suit Mia. She adjusted some of the features in the botspace and tried again! The new suggestions were wacky and fun.
In video 3, Mia's intention is to limit the digital distractions in her work environment and her weekend getaway. She assigns taskmaster bot to prioritize this intention in her botspace.
INSIGHTS
After a few iterations, pairing a bot team to an intention, users proposed combining several intentions with their bot teams to create what we called a super-superbot.
All users interviewed preferred the multi-bot approach over monolithic AI.
Some users preferred their AI to be more autonomous while others preferred to micro-manage. All users desired the ability to adjust the level of management over time.
Users found the idea of an AI they can see inside less threatening than an opaque AI.
Users found the concept that the AI could be “only what you put into it” empowering.
AI Considerations: There are several opportunities in the AI design space for this work, such as training for data prioritization and intention with multiple bots in an internal system, as well as combining bots from multiple systems to engage in decision making. The bot experience would differ regarding the type of learning (supervised or unsupervised), and access to data.
Real World Challenges: While this is a playful approach focused on increasing intention in work and relationships, there are some interesting real world challenges in this space. Questions that emerged from this research include:
What does it mean to be granted limited access to data from a loved one, and what are our changing expectations regarding privacy and our personal portable AI information and knowledge?
How might this intelligent experience change when there is a lack of data from one or more parties in a collaborative bot space?
If we incorporate this bot decision making in collaborative spaces, how might we inform users that the decision was made by a bot?
How can we visually understand the decision making process between all bots and users?
LEARNINGS
Through control and transparency there is an opportunity for increased trust in AI systems. We believe that we can use bots:
To Mediate, Manage, Communicate
As Self Expression
As Means of Empowerment
As A Vehicle for Collaboration
Industry: Smart Assistants, Bots
Expertise: Cultural AI Design, Experience Design, Intention Design
Project Date: September 2016 - December 2016