Socius: a smart cities project

An academic research project investigating how icon design affects donation decisions in the users of an experimental direct-donation platform.

The Socius Research Group is an NSF-funded research initiative. Min Kyung Lee is the primary investigator.

TEAM | Min Kyung Lee, Jitae kim

ROLE | Design Researcher

TOOLS | Qualtrics, MTurk, JMP

METHODS | literature review, experiment design, statistical analysis

Socius is a collaborative research initiative that works on issues of homelessness, food insecurity, and resource allocation in the city of Pittsburgh.

Socius is a collaborative research initiative that works on issues of homelessness, food insecurity, and resource allocation in the city of Pittsburgh.


What if you could donate directly to people in need in your own community, giving them exactly the things they need to feed their kids or clean their homes? The Socius Research group has designed an experimental direct-donation website that allows users to do exactly that. After collecting data about needed items from food pantry users, the website displays what items are needed in what location. Users of the website are then able to select items they wish to donate and have those items delivered directly to the food pantry.

However, many questions remain about how to best design the direct-donation website to encourage users to donate. In this project, we designed a series of experiments to test the effects of different icon designs on willingness to donate (WTD) in the users of the website.

    My role in the project was to conduct a literature review and, based on the literature review findings, design and implement pilot and manipulation-check studies looking at different icon designs and possible moderating factors in users' donation decisons.


    Research approach

    current work

    Pilot Study: Icon Effects | We also used this pilot as a chance to test a wide variety of icon to look for hints about what styles of depiction might be worth testing further.

    Icon Manipulation Check | Tested whether icon designs were perceived as intended. If an icon was intended to be a 40-yr old woman, for example, we asked participants to select the icon's gender and age.

    Study 1: Identifying Information vs. Depiction of Need | Here we tested icons that conveyed varying levels of identifying information (such as age, gender, etc.) against icons that depicted the item being requested or the need for the item itself. 

    Future work

    Study 2: Array vs. Aggregate for Groups | Groups of individuals can be represented either as an aggregate (with size or density or some other attribute representing how many individuals are represented) or as an array, with each individual represented as a single icon in a large group of icons. In this study, we propose to test aggregate and array representation of individuals in need and to measure willingness-to-donate. 

    Study 3: Numeric Representation vs. Non-numeric | In our early pilot tests, we found that showing participants the number of items requested appeared to have a positive effect on willingness-to-donate. Here, we propose to follow up on that finding with an experiment that compares numeric and non-numeric icon designs.


    Key Concepts from Donation Research

    identifiable victims vs. statistical victims

    The most prevalent finding in the literature review was that potential donors were more willing to donate when they were given identifying information about the recipients of their donations--things like name, gender, age, or profession. Most strikingly, even very small changes in identifiability changed willingness to donate. Knowing a sick child's name or age, for example, made potential donors much more willing to donate. However, even greater identifiability--such as showing a picture of the child--increased likelihood and amount of donations even more (Kogut & Ritov). 

    However, there are drawbacks to using the identifiable victim effect as a solicitation strategy. In practice, revealing identifying information about an individual often leads nonprofits to use representative individuals, which can be deceptive to donors. This NYTimes article gives a nuanced look at that practice. 


    psychological proximity + target specificity 

    Similarly, several studies have found that potential donors are more willing to donate to recipients who are psychologically proximate to themselves (sharing a gender, location, profession, etc.) and are more willing to donate to individuals that groups. For example, Galak, Small, and Stephen found that lenders favored individual borrowers over groups of borrowers, a pattern consistent with the identifiable victim effect. They also favor borrowers that are socially proximate to themselves. Across three dimensions of social distance (gender, occupation, and first name initial), lenders prefer to give to those who are more like themselves.

    And, as with the identifiable victim effect, there are significant downsides to using psychological proximity as a solicitation strategy. Because of homophily effects (which find that we tend to empathize with and help those who are similar to us), giving identifying information about recipients can reinforce inequality, since those in socially advantaged groups will be more likely to donate to others of their group.

    guiding questions

    Disclosure: how much information should be provided about recipients?

    Should the focus be on the person or on the problem/need?

    Level of visual finish: polished or handmade-looking?

    Hint from the pilot: will numbers increase WTC?


    pilot study

    Will varying the amount of identifying information given by icons affect donations or attitudes in potential donors?

    In this first study, we aimed to pilot a wide variety of icon designs and check for several different moderating variables. Fundamentally we were interesting in finding out what kind of icon willingness to donate and desire to help, and so these were our dependent variables. However, we were also interested in in exploring possible reasons why a given icon performed better or worse than the control, so we also asked participants if they felt sympathy for the person depicted, if they felt sad after viewing the image. if they felt similar to the person depicted, and if they felt the person depicted was deverving of help. 

    Study materials

    The image you see is part of a new website that will allow users to donate items that have been requested by food pantry users directly to food pantries. 
    This icon on this map represents a real food pantry user and the items that they requested.
    Please imagine that you are using this website. Take a look at the image and then answer the questions below.

    Study Conditions



    Two dependent variable measures: “How much would you donate?” (given in dollars), and “How much did you want to help the food pantry user?” (given on a 7-pt likert scale). 

    Moderating variable measures: “I felt sympathy for the person depicted”, “After viewing this image, I felt worried, sad, or upset”, “I felt similar to the person depicted”, “the person depicted is deserving of help” (7-pt likerts). 

    60 participants, 13 icons

    On MTurk, HIT required 95% approval rating, over 100 HITs completed, live in US, and over 18.



    We were surprised by how much higher the mean donation amount was for the basket condition than other conditions, and planned follow-up up study to evaluate icons the depicted need or action, rather than the human beneficiaries of the action. 

    Here we saw higher levels of willingness-to-help for both the basket icon and the drawing icon (which looks like a child's self-portrait). We were surprised by the success of the hand-drawn icon, and planned a study to test whether hand-drawn icons motivated greater donations or whether this effect was because the icon looked like a child. 


    manipulation check

    This manipulation check experiment was designed to test whether our icon designs communicated as they were intended to. We checked age, race, gender, emotion, and attractiveness—all factor that we thought might affect donation decisions—to make sure they were consistent across similar conditions. 

    After considering the results of the pilot study, we added additional icons to tease apart confounding effects. In our hand-drawn icon condition, for example, we added a drawing of an adult woman to separate the effect of depicting a child from the effect of showing a hand-drawn icon. 


    Desired outcomes

    All human conditions that show a facial expression be perceived as sad. ✔

    Photo and demographic icon conditions be perceived as female and as the intended race. ✔

    Photo conditions be perceived as similarly attractive and of a similar age. ✢

    Ambiguous condition be perceived as varied or indeterminate race, ethnicity and gender. ✔

    Hand-drawn icons perceived as female; one adult, one child. ✔


    All icons tested were perceived as intended except for the photo conditions. There, we found significant disparities between the two photos' perceived age and attractiveness. We began exploring pre-tested databases of photo portraits to find alternatives photo icons.

    Finally, we also asked participants why they chose to donate (or not). In reading through these answers, we observed that racial bias appeared to play a role in donation decisions. This confirmed the importance of homophily effects, and underscored the need to explore non-identifying ways to represent people in need.