The Story of Chore Wheel
Reflections on the creation of a "new system of governance"
This month marks two years since Chore Wheel was first developed for use at Sage House. Initially an experimental system, Chore Wheel has matured into a daily tool enabling multiple generations of housemates to effectively share space.
This essay will look back on the creation of Chore Wheel, its launch, adjustments and pivots made over the first two years, and where the project might go from here.
Imagining Chore Wheel
The seeds of what would ultimately become Chore Wheel were planted in late 2016, beginning with my voting systems research as a graduate student at Columbia. My thesis, An Analysis of Pairwise Preference, was a broad investigation of the possibilities of “pairwise preferences” — a single choice between two options — as building-blocks of more complex voting and decision systems. The work was well-received, earning an A+ and a profile on my department’s website.
After graduation, I put the idea to the side, focusing on my new job as a machine learning engineer at Foursquare.1 However, my father, a property manager and housing advocate, would miss no opportunity to encourage me to consider a career in housing. While initially dismissive, over time I began to see ways my more recent academic interests might combine with my earlier experiences with housing co-operatives to create something truly innovative and impactful.
In May 2018, after a year-and-change at Foursquare, I left New York, moved abroad, and got a job as a research engineer at Colony, developing experimental voting and budgeting systems. Having been interested in web3 for a while, I jumped at the chance to work in the field.
It was a good fit. Within a few months of joining I was invited to London for a “research summit” where, along with my colleagues Jack du Rose and Aron Fischer, we developed BudgetBox, an early public-goods-funding mechanism.2
BudgetBox combined the computational techniques initially developed for pairwise preferences with new ideas of continuous funding emerging out of the DAO ecosystem to create a complete mechanism for collectively managing shared resources. Ultimately, BudgetBox would go on to form the core of Chore Wheel’s chores system.
Web3 is full of people with revolutionary zeal, hoping to use their new technology to reform world governments and transform global economies. I felt differently, wanting to explore ways these tools could be used locally, solving everyday problems for regular people.
With frequent conferences in Europe, there were many opportunities to meet and mingle with cypherpunks in Paris and artists in Berlin; it was a heady time of pre-pandemic exploration and discovery, and of gathering influences: around interface designs, reputation systems, engineering techniques. I had big plans, and wanted to get things right.
Developing Chore Wheel
In summer 2019 a friend sent me an article about a coliving startup doing interesting things with technology; it felt like the moment had arrived. I had a rough plan: buy and renovate an old house, and use it to pilot this experimental governance system.
In February 2020 I pulled the trigger and moved back to my hometown of Los Angeles, where I thought my odds of success would be the highest. What I hoped would be an exciting adventure quickly became a harrowing ordeal, as the pandemic quickly set in. “Alea iacta est,” I said, and threw myself into the project.
At this time, I entered into ongoing collaboration with two friends: Seth Frey, a professor and computational social scientist, and Joseph DeSimone, a brilliant game designer. I knew I needed feedback on my ideas, and recruited them to advise me.
From 2020 - 2022, we refined the mechanisms that would form the core of Chore Wheel and wrote the Mirror Whitepaper.3 It was a productive and engaging collaboration; I often joke that I came into our work with most of the mechanisms fairly well developed out, and that Seth and Joe’s job was telling me which of my ideas were the least bad. As is often the case, apparent simplicity often conceals substantial complexity. Through an extended series of letters, we worked through many possible scenarios and circumstances. There was something charming and “Federalist Papers”-esque about our collaboration: discursive epistolary explorations of politics, power, fairness, and freedom. We had fun.
One significant piece of advice that Seth gave was to begin as simply as possible. To paraphrase, “if you start with many interacting mechanisms, it will be difficult to figure out why things are going wrong. If you start with only the essentials, it will be a lot easier to refine and expand the system.”
With that in mind, we planned to launch with what I saw as the three core mechanisms of Chores, Hearts, and Things.
Launching Chore Wheel
In September 2022, the month Sage opened, I started implementing Chore Wheel in earnest. My goals from a software perspective were to create something that was auditable, robust to error, minimally complex, and adaptable.4
The plan was to implement the tools as Slack bots. As discussed in our research paper, we drew explicit influence from Discord’s gaming communities, who have long used chat-bot integrations to perform internal governance tasks, such as sorting members into raiding groups or helping victory parties fairly divide in-game rewards.
It was a busy month; I had to implement three separate products while simultaneously navigating the in-person experience of Sage House. It was also a special time: three eager new housemates having the run of a big beautiful house, enjoying the novelty of the experience and looking forward to good times ahead.
The “bare essential” functionality was still a long list: the ability to define, value, and claim chores; the ability to update group priorities; the ability to “take breaks” and reduce one’s obligation; the ability to vote on claims; the ability to initiate and resolve conflicts; the ability to give good karma; various automated “heartbeat” features; and so on.
On balance, implementation went smoothly. I had set a goal of launching Chore Wheel before our fourth housemate arrived in October. After a few weeks spent implementing, and a few days configuring Amazon Web Services, we were live.
Iterating Chore Wheel
In some ways, launching Chore Wheel was like putting a rocket into space: a complex endeavor with many moving parts, many factors beyond my control, and very few chances to try again if things didn’t go right. If the system failed, it would be very difficult, if not impossible, to get the buy-in to try again.
Fortunately, whether by luck, skill, or mystical synchronicity, everything worked more-or-less right out of the gate. People figured out how to claim chores, understood the idea of earning points, and got comfortable using Slack to communicate. There was very little resistance to using a new system, folks clearly saw the value, and there were relatively few bugs in the experience. To say I felt relief would be an understatement.
That said, there was ample room to grow. New housemates frequently had suggestions on how to improve the system, often challenging my own assumptions about that Chore Wheel should or shouldn’t be. Helping housemates develop these ideas and integrating them into the tools became an ongoing process:
Adaptations and improvements
Chore gifting
One of the earliest changes we made was adding support for “gifting” chore points. Making points transferable achieved two things: first, it made it possible for people to “split” chores, and second, it made it possible for people to gain points for making contributions outside of the constraints of the pre-defined chores list.
Chore editing
A more complex change was adding the support for housemates to edit the chores list themselves. Initially, the list was updated manually, with no way for housemates to make changes on their own. While this was done initially out of expedience, over time it became clear that a proposal-editing flow would allow me to shift more decision-making to housemates and ultimately increase the legitimacy of the system.
Ongoing explorations
Points minting
A frequent request was to make it possible to “mint” points via a vote, in order to recognize special or one-off contributions. I have been resistant to such requests, on the grounds that this would throw off the balance of the points economy, in two ways.
First, a key concept behind the chores system is that the amount of points available is fixed, at 100 points per person per month. Points are not a currency to be accumulated, but an ephemeral token for contribution accounting. The fixed amount of points allow for a continuous flow “from source to sink” through which points accrue in chores, are claimed by housemates, and evaporate at the end of each month.
Point minting would imbalance this “steady state” by introducing more points into the system than people need to meet their obligations.
Second, a concern is that point minting would introduce a political dimension to the chores system which is currently, and thankfully, absent. If point minting were an option, then the door would be open for housemates to request points for arbitrary contributions, introducing significant mental and emotional labor in evaluating and discussing the relative merit of this vs. that contribution.
None of these issues are insurmountable — as long as point minting was offset by a commensurate reduction in continuous point accrual, and the system’s steady-state was preserved, then in principle this functionality could add a useful layer of nuance.
Chore prioritization
The distributed chore prioritization system is Chore Wheel’s most innovative component. The underlying process, by which pairwise preferences are aggregated to create a total prioritization over all chores, has been largely unchanged since launch. What has changed, and changed often, is the interface around the mechanism.
I believe — and have believed for many years — that the mechanism of resource allocation via pairwise inputs is an innovation with paradigm-shifting potential. The biggest challenge at this point is iterating and refining the interface to the point where casual users can both intuitively grasp and fully leverage the mechanism. Getting there in the context of Chore Wheel has been an ongoing process:
Initially, all preference updates had to be one pair at a time, discouraging housemates from making larger changes. An update was made allowing many pairs could be submitted at once.
Initially, less engaged housemates were given “implicit preferences” which limited the influence of more engaged housemates, a well-intentioned choice which ultimately discouraged more engaged housemates from participating. An update was made reducing the implicit preference, such that individual housemates have more noticeable impact.5
Initially, preferences could only be expressed as the binary “all or nothing.” An update was made allowing for more levels of preference: mild, corresponding to a 70% preference, and strong, corresponding to 100%.
Initially, housemates would not see the effect of their updates until after submission, creating a confusing experience. An update was made providing a “preview” screen showing the effect of the proposed change, allowing housemates to iteratively revise their update prior to submitting.
The preview screen at the end of the prioritization flow Initially, the mechanism was presented in a formal and academic tone, which was off-putting to some housemates. An update was made presenting the preference updating flow in a more narrative style, intuitively guiding housemates through the decision process.
The Future of Chore Wheel
After two years of continuous operation and with over 2,000 chores claimed, it is fair to say that Chore Wheel works. Looking forward, there are two main opportunities:
Growing the base. The work now is to learn how to reach more communities, groups who struggle with domestic issues but don’t know about Chore Wheel or can’t envision how it might become a part of their day-to-day. More communities using Chore Wheel means more perspectives to help inform development.
Adding nuance. While the fundamental mechanisms of Chore Wheel are proven, there are often situations which call for more flexibility and nuance. Adding this subtlety will make Chore Wheel more useful for more people.
Creating Chore Wheel has been a journey of many years and a labor of love, and I look forward to seeing where the journey goes.
You can learn more about Chore Wheel and try it for yourself here.
During this time I developed Talmud.ai as an early study of a pairwise-preference oriented interface design.
“Mirror” was the original name for Chore Wheel, evoking the way that the software merely “reflects” the views and values of the community.
Drawing on the functional programming influences I had acquired at Foursquare, Chore Wheel was designed as a series of append-only logs and functional operations on those logs, supplemented by more mutable configuration tables for storing infrequently-updated parameters. A high-level goal was to make it impossible for the system to ever enter into an “inconsistent” state, and to be able to recover gracefully from shutdowns. For the UI, Slack’s Bolt framework was easy to use, some clunky design choices notwithstanding.
We could make a hand-wavy information theory style argument that I had initially over-estimated the information content of the “implicit votes” of less engaged housemates, and simply corrected for this later on.