%!PS-Adobe-3.0



Social Engineers: Enabling the Matchmaking Game in Online Dating Systems

Shelly Farnham

Social Computing Group, Microsoft Research

One Microsoft Way

Redmond, WA 98052-6399

shellyf@

(425)706-6394

Jens Riegelsberger

University College London

Gower Street London,

WC1E 6BT, UK

jriegels@cs.ucl.ac.uk

+44 207 679 3643

ABSTRACT

In creating online dating systems, researchers and practitioners must focus on how to best design a system to increase the likelihood of two people finding each other and having successful dating experiences. At the core of online dating is the profile, through which people convey information about the self, and through which others filter for those with whom they want to initiate a social interaction. We found in two studies that the presence of photos in profiles had a negative impact on the overall ratings of attractiveness of the person profiled. However, photos provide important information through which people will make successful matches, so users should be encouraged to provide photos in their profiles.

INTRODUCTION AND BACKGROUND

In the past few years, online dating has become increasingly main stream. Millions of people use matchmaking tools such as and Friendster, and online communication technologies are increasingly integrated into people’s day to day social relationships [2].

Generally, online dating systems provide impression management tools that allow users to project their identity into a large directory of candidate dating partners. At the core of online dating is the profile, through which people convey information about the self, and others filter for those with whom they want to initiate a social interaction. Users then browse through these directories of people’s profiles, often filtering people out based on features such as geographical location, age, sex, and level of education.

The role of researchers and practitioners in designing online dating systems is much greater than just that of creating usable user interfaces. The goal of online dating and matchmaking applications is to engender positive, concrete social outcomes such as the development of face-to-face friendships and romantic relationships. Thus as researchers and practitioners create online dating systems, they must take on the role of social engineers: how can they best design a system to increase the likelihood of two people finding each other and having successful dating experiences.

For example, people generally achieve greater levels of intimacy with others through graduated levels of reciprocal self-disclosure [1]. Thus an effective online system should provide tools for gradual transitions between people who have observed each other’s profiles, to people who send each other semi-anonymous messages, and then through their own personal, completely identifiable email messages. Friendster, for example, allows people to send private messages within the system. provides tools for people to email each other with an email account generated .

Social engineers of effective online dating and matchmaking systems must take into consideration basic principles of attraction. One primary determinant of attraction is similarity of personal characteristics. (The stereotype that opposites attract is not generally supported by studies of interpersonal attraction[1].) Thus an effective matchmaking system would match people based on demographic characteristics, similarity of interest, values, and so forth as provided in the users’ profiles.

Geographical proximity is another predictor of attraction, because geographical proximity (e.g., living in the same dorm) engenders greater frequency of exposure, and people tend to like that which is familiar. Thus online systems should provide tools for frequent exposure and interaction with the same person. Evite, for example, provides tools through which people may observe whom in their social circle will be attending any particular event.

People also tend to like others that people they like also like (called balance theory in the field of psychology [1]). That is, if Jane likes John, and John like Mary, then Jane is likely to like Mary. Many recent applications have been taking advantage of this balance tendency, by placing profiles in a social context. Friendster, for example, has people link to their friends’ profiles, so that other users may navigate around their social space, increasing the likelihood that they will find someone they like.

Perhaps most importantly, people tend to prefer physically attractive others as dating partners. We might expect then that the presence of photos is a very important predictor of liking for a particular profile. However, contrary to predictions, we have not found this to be the case.

STUDIES OF PHOTOS IN PROFILES

Several years ago we were interested in exploring the impact of photos on preferences in online dating. At the time we had observed that people were often hesitant to post photos about themselves in online dating systems, for fear of the stigmatizing effect of being “found out” by their friends and coworkers. Our goal was to show that the presence of photos was an important predictor of people liking the profiles. We conducted a simple online study, where we varied the presence or absence of photos in profiles, and then had study participants rate the profile on a number of personality dimensions. At the time we found, contrary to our prediction, that people appeared to like profiles less in the presence of photos. Given that we did not predict this outcome, and could not readily explain it, we did not pursue the topic.

However, a couple of years later we recently conducted a similar study [4], examining the impact of media types on preferences in an online gaming scenario. For this study we had XBox live players create profiles in a study administered through our Online Lab [3]. Each participant created either a profile with just text, a profile with text and a photo, or a profile with text and a voice clip. We then had another set of 267 (96% male) participants rate 25 profiles in each condition (text, text with photo, text with voice) on the extent to which they might want to play a game with that person. Ratings were aggregated across the profiles for each individual.

We found that people preferred profiles less when photos were present, although they preferred profiles equally well if they listened to a voice clip of the person profiled. See Figure 1. Participants wanted to play with the person profiled less (F(1, 171) = 8.26, p = .01), they expected the conversation to be less enjoyable, (F(1, 171) = 7.18, p = .01), and they expected a less satisfying game (F(1, 171) = 7.34, p = .01), when evaluating profiles with photos than when evaluating profiles without photos.

An examination of the distribution of ratings for wanting to play with the person suggests that people used photos to be more discriminating in their decisions. Compared to text profiles, the distribution of responses for photo profiles shows a peak for very negative responses. Thus, photos seem to work as a filter: people like them because they help them decide whom they do not like.

[pic]

Figure 1: Photo profiles were consistently rated lower.

[pic]

Figure 2: Photo profiles show a peak for negative responses, where 1 = I do not at all want to play with this gamer.

The replication of this effect, that photos had a negative impact on ratings, inspired us to go back and reexamine the results of the previous profiling study. In that study, 277 participants (222 males, and 53 females) had completed a simple dating choice task, also administered through the Online Laboratory.

Participants were presented with two profiles, with the instructions that they were to examine the profiles carefully, and then select which person they would most care to date. We had created five male and five female profiles, adapted from profiles found in , and randomly combined two of the profiles depending on the study participants’ specified gender preferences. We then had the participants rate the personality of the first person profiled on a number of dimensions. Participants were randomly assigned to see the profiles either with or without the presence of photos.

We then factor analyzed the ratings of personality dimensions to assess which loaded (>.5) on a positivity factor, and then summed these ratings into one overall positivity rating. We found, as mentioned earlier, that people generally rated profiles in the presence of photos less positively, F(1, 249) = 5.96, p < .02). See Figure 3.

[pic]

Figure 3: Photo profiles were rated as less positive.

However, we also found that participants appeared to be able to form somewhat more complete impressions of profiles in the presence of photos. For each personality dimension, we provided people with the option of declaring they “do not know” how to rate the profile on that dimension. People selected the “do not know” option more frequently in the presence of photos (M = 1.35) than in the absence of photos (M = .99), t(251) = 1.65, p < .05, 1-tailed. Furthermore, people seemed to want photos. Following the task of rating the profiles we had people suggest improvements for the profiles. 27 out of the 139 people, in the no photo condition, 19%, spontaneously suggested we add photos to the profile.

Again, these results suggest that photos, while decreasing the positivity of the impression formed about the person profiled, increase the discriminatory abilities of the person making selections from the profiles.

CONCLUSIONS AND DISCUSSION

We found in two studies, contrary to our predictions, that the presence of photos had a negative impact on the overall impression of profiles.

Although one might be tempted to suggest to users that they not post a photo, because odds are they will be less liked, we must remember or role as social engineers, where our ultimate goal is a successful match.

Going back to basic psychological principles, although people tend to prefer physically attractive others, they tend to successfully match with people who have similar levels of attractiveness--known as the matching hypothesis [1]. Classroom simulations of this effect—called the matching game--illustrate that it occurs largely through an iterative selection process, irrespective of any person’s awareness of his or her level of attractiveness. That is, as people attempt to partner with the most attractive person possible, they will make bids for partnerships with the most attractive potential partners first, who then accept the highest bidder. Obviously outside the classroom or experimental studies attractiveness is comprised of a much more complex combination of features, of which physical attractiveness is only one. However, photos are clearly an important discriminator in online dating systems. Photos allow users to filter out those whom they would not care to date.

Thus as social engineers, we should provide tools in online dating systems for a similar process of elimination, allowing people to find others with similar levels attractiveness. Reputation systems, or other indicators of social standing, for example, might aid users in finding others with whom they might be compatible.

That the presence of photos might negatively impact a profile’s attractiveness highlights the fundamental tension between the desire to provide a positive impression of the self to others (which may entail not posting a photo, misrepresenting one’s age, income level, etc.,), and the desire to find the most attractive partner possible (thus wanting to see photos, and have accurate information in profiles). As social engineers, we need to develop matchmaking tools with the realistic expectation in mind that a person’s own level of attractiveness is the best predictor of whom else will prove to be a successful match.

REFERENCES

1. Baron, R., & Byrne, Do. Social Psychology, 8th edition. Allyn & Bacon: Boston.

2. Farnham, S., Kelly, S., Portnoy, W., & Schwartz, J.L.K. (2004). Wallop: Designing for Co-located Social Networks. In Proceedings of HICSS-37, Hawaii

3. Online Laboratory: .

4. Riegelsberger, J., Farnham, S., Phillips, B., Lee-Tiernan, S. (2004). Personality Matters: Rich Modalities for Matchmaking in Online Games. Unpublished manuscript.

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download