Aliyah Banerjee, The University of Texas at Austin
Safety apps have become popular solutions to address women’s safety in India, using features like panic buttons, location-sharing, and SOS alerts to enable safe travel at night and aid in instances of sexual violence or harassment.
However, a lack of digital literacy, feelings of constraint with location-sharing services, fear of control by abusive partners, data privacy concerns, and perpetuation of sexual violence stereotypes pose significant barriers for female users of these apps. To alleviate these concerns, using one-sided matching is proposed to build a novel safety platform for women.
Design
Since users dislike constant monitoring by family members and sharing location details with developers, an alternative is directly matching women using the app with one another, forming “virtual buddy” pairs for mutual safety through periodic check-ins. This approach resembles the “stable roommates” problem in matching theory, introduced by Gale and Shapley, where a set of 2n elements (roommates) must be partitioned into n pairs.
When going out, users log on to the app and see a list of other users who are looking for a buddy at the time, ranked by criteria such as age or location. The user then “proposes” to their first-ranked potential buddy, and if the other user accepts, they form a virtual buddy pair and are shown to be “unavailable” on the app (so that a matched user cannot abandon a pairing for a new buddy). If the proposal is declined, the user proposes to their next-highest ranked buddy and so on until their request is accepted. The user simultaneously responds to proposals from other users. The length of this matching process is dependent on factors such as the number of current users, the amount of time a user is willing to spend on the app looking for a buddy, and the extent to which a user desires a virtual buddy. To focus on the matching process, intentionally matching with a long-term buddy or scheduling partners in advance is outside the scope of the app.
Within a matched pair, the app will initiate a short phone call using the phone’s calling system, allowing the users to check in with each other about their safety situation. Either user can press an SOS button to alert authorities should their buddy not respond to numerous calls. Authorities can then directly contact the distressed individual or alert family if the user has provided that information. A publicly-displayed rating for each user may be implemented, based on feedback from previous buddies, discouraging dishonest usage of the app’s features and promoting speedy responses.
This simple design ensures access for digitally-inexperienced users and control over sharing sensitive/location-related data while still having a trusted check-in buddy. Users are encouraged to use the app whenever they feel uncomfortable venturing out alone, promoting an independent approach towards safety that women lack due to victim-blaming.
Desirable properties
Stability (individually-rational without blocking pairs) is a well-established desiderata in matching theory to evaluate the quality and durability of a given match. How well do virtual buddy pairings satisfy these properties?
Individual rationality is satisfied by a matching when no person prefers being unmatched to being with their current partner. Any possible matching of users on our application has to be individually rational, because a user must accept another user’s proposal before they become virtual buddies.
A blocking pair is formed when two individuals (who are each matched to someone else) prefer each other to their current partners. Blocking pairs may threaten the stability of existing buddy pairings in the case of sequential pairings. Users A and B may be matched into a buddy pair. C and D come online later and are also matched. However, A and C prefer to be matched together than with their current buddies, had they been online at the same time.
To tackle this challenge of unfulfilled stability, Atay et al.’s concept of ‘weaker stability’ can be used, which states that an individually rational matching is ‘weakly’ stable if all possible blocking pairs are also “weak.” Weak blocking pairs for the original matching are those that lose their credibility if another blocking pair is formed for the original matching with a more preferred partner. Weak stability, like stability, relaxes the notion of virtual buddy pairs being less likely to break.
Formation of future blocking pairs when more preferred users arrive undermines the credibility of previously existing blocking pairs. If two users form a blocking pair, and one of them prefers another user more who enters later, then the original blocking pair loses its credibility. This process could continue indefinitely with new users entering the application. Since all blocking pairs eventually lose credibility and all pairings are individually rational, any buddy matching satisfies weak stability. This also eliminates the danger of pairs breaking and female users in India being left alone in unsafe situations.
Limitations
A complete preference ranking of a user over all other users simultaneously online, as required in Gale and Shapley’s roommates problem, is impossible to work with due to the ever-changing pool of arriving users. Users may also withhold their location, age, and other characteristics from the application, leading to them being excluded from rankings of potential buddies who prioritize those criteria in a buddy.
Matched users also cannot be prevented from abandoning a pairing, except through restricting them from seeing other online users until both buddies indicate their outing is over. Users may still exit the app and their pairing, re-entering as a new user.
Matching theory has great potential to improve existing women’s safety apps in India. Following such a design, it is possible to leverage technology to provide safer options to Indian women of all digital literacy levels while alleviating privacy concerns and giving users control over their own safety. While the potential is clearly present, advertisement for widespread usage, preference ranking of users over potential buddies, proper regulation, and refinement of the design itself are areas for further development.