Introduction
The division of household labor is a persistent source of relationship conflict. While a simple 50/50 split is often the goal, it frequently fails because chores are indivisible and individuals have varying preferences. In computational social choice, chores are viewed as items with "disutility" (negative value). The challenge lies in finding an allocation that is both Pareto-optimal (PO)—where no one can be better off without making another person worse off—and Envy-Free up to One Good (EF1)—where no one envies another's load after hypothetically removing a single chore.
Methodology: Quantifying Disutility
The Kajibuntan app bridges the gap between theoretical fair division and real-world application by simplifying preference elicitation. Users do not directly assign numerical disutility; instead, they provide two intuitive metrics: Time ($t$) and Preference (Like, Neutral, Dislike). The system then calculates total disutility as $t$, $2t$, or $3t$ based on these inputs.
The core engine uses a modified Discrete Adjusted Winner (AW) Algorithm. Unlike standard protocols, this version randomizes the starting "winner" and "loser" to improve perceived fairness. It continues transferring chores until an EF1 state is reached, stopping only if a transfer would create envy or widen the disutility gap.
Visualization and User Agency
A key feature of Kajibuntan is making "invisible" labor visible. The app uses pie charts where slice sizes represent the proportion of total disutility rather than just the count of chores. This provides an objective baseline for negotiation.
Furthermore, the app allows for manual adjustments. Users can "swap" chores by clicking on chips, combining algorithmic rigor with human agency to ensure the final result feels "right" to both parties.
Results and Impact
Following a launch on Japanese national television (NHK), the app saw significant engagement with over 13,000 unique visitors in its first month. A survey of 131 users revealed that 60% found the tool useful for facilitating discussions. This indicates that the app's primary value is not just the mathematical "answer," but the structured framework it provides for communication.
Limitations and Future Work
Current limitations include the assumption of additive disutility, which ignores combinatorial effects (e.g., the efficiency of doing related chores together). Future research will explore empirical perceptions of fairness to better align mathematical definitions like EF1 with human psychology and investigate potential strategic behavior in preference reporting.
Summary updated