Cities

Center fоr Interacting Urban Networks

A Collaborative Real Estate Search Tool

Student: Junior Francisco Garcia
Supervisor: Azza Abouzied

Existing real estate search tools like Airbnb enable users to look for properties based on a variety of interest points such as desired location, price range, or availability. Although such tools are able to return results that satisfy the criteria specified by a single user, little work has been done to extend the tools’ capabilities to provide explicit support for multi-user collaborative search. Real estate search is an activity that oftentimes involves more than one stakeholder. Groups of friends going on a vacation or families looking for a new place to live find themselves at a crossroads whenever they need to compromise their preferences when looking for the place they will inhabit for a significant period of time. Finding a suitable compromise in this scenario is challenging albeit crucial, as one or more users in the group may find themselves living in and paying for a place they dislike.

The goal of this research project is to investigate users’ search behavior in order to create a system that can support collaborative search and find optimal compromises. An extensive literature review was carried out in order to study existing collaborative search tools and understand how they were able to enable collaboration amongst different users effectively. This understanding of collaborative search is applied in the context of real estate, a search topic that requires users to think numerically, qualitatively, and spatially in order to obtain a result that they are satisfied with. A single search query may take into account the landmarks a user wants to be in close proximity, the price range the user can afford, whether or not the property is in a school district, the size of the property’s lot, and the aesthetic qualities of the property. Given how complex a single real estate search query might be, are there new search metaphors that could be used in order to best settle diverging or conflicting preferences amongst searchers? Even within the single user search scenario, are there other ways to formulate real estate search queries that could prove to be useful when enabling collaborative search? Are existing real estate search tools missing ways searchers think of space, geographical data, or living preferences when implementing their search engines? In a collaborative real estate search tool, these questions are equally as prevalent and are further magnified by the social dynamics at play when a non-trivial decision needs to be made by a group of people. Are a person’s preferences more important than the preferences defined by their group members? How can a tool balance the importance given to a specific preference? Is it even the tool’s responsibility to assign such importance and it instead should mediate the conflict resolution process if there is any? We hope to answer these questions through an innovative “personal-shopper” user study that groups people of different personalities together and asks them to use our tool to find a property.

As part of a capstone project that culminates in May 2021, we are creating a human-in-the-loop search interface that employs 3 different conflict resolution mechanisms: swap-to-settle, hybrid tags, and a global ranking of results. Swap-to-Settle allows User A to offer alternative properties to User B in exchange for a property User B defined. Hybrid tags allow users to define new tags not originally supported by our system that might solve conflicts (i.e. a room with a balcony as a way of solving a smoking-asthmatic conflict). And our global-ranking algorithm will rank the properties selected by the group by taking into account likes, dislikes, and the satisfaction of the combined preferences defined by the group. These three conflict-resolution mechanisms will then be tested through a large-scale user study involving 5 groups of 4 people each. Extensive qualitative and quantitative analysis of the user study’s results will be carried out in order to draw conclusions and extend the literature on collaborative search.