Humanizing Algorithmic Matching in an Online Dating App

Abstract

When building online systems designers have to make many decisions that impact user experience. One of said choices is whether certain interactions are automated or remain human. To make matters more complicated, they can also decide to make certain interactions seem automated or human. Making the right decision has increased in importance as more of our interactions are taking place online. Context The aim of this thesis is to discover if guidelines can be developed to facilitate designers when making these decisions. It does so by taking a new dating app called Breeze as a use case. This dating app is different compared to existing dating apps because it automates the date arrangement process. Instead of being able to swipe through profiles and chat with matches, users receive two profiles per day and directly fill in a date picker when matched. Breeze then arranges the date and lets the match know where to be at what time. The app faces the problem that many of the dates get canceled because users stop responding during the date arrangement process. User research reveals that, by having limited online interaction, this process has become impersonal, anonymous and inflexible which makes it easy to dehumanize your match. Dehumanization leads to loss of commitment when arranging a date. This leads to the research question: What would the impact on commitment and the overall user experience when this process would instead be humanized? And how can these learnings help designers of other online systems decide when and how to humanize interaction? Theory To answer these questions, this thesis first elaborates why dehumanization is inevitable when interaction is mediated by technology. Dehumanization can be mitigated by humanizing interaction. However simply humanizing interaction is not the solution because, apart from its advantages it too has disadvantages. The ideal balance between dehumanization and humanization depends on the type of platform in question. Findings Through conducting two Build-Measure-Learn loops, this thesis finds that the ideal balance does not only differ between platforms but also differs within the customer journey of one platform. Within each step of the journey, users have different interaction needs, which not only depends on why they decided to interact with the system in the first place, but also with whom they are interacting at that point. Additionally the companies that build these systems also have conflicting needs, which depend on their strategy and available resources. Result In order to find this ideal balance between humanization and dehumanization, this thesis proposes the Framework for humanized interaction. This framework is validated through conducting expert interviews, a pilot with an external company and by applying it to the use case. The latter results in new concepts and recommendations suitable for implementation by Breeze.