SELF-DISCLOSURE AND PERCEIVED TRUSTWORTHINESS OF AIRBNB HOST PROFILES Xiao Ma [1] Jeff Hancock [2] Kenneth Lim Mingjie [3] Mor Naaman [1] [1] Social Technologies Lab, Cornell Tech [2] Department of Communication, Stanford University [3] Department of Computer Science, Cornell University xiao@jacobs.cornell.edu | maxiao.info
WHO DO YOU FEEL MORE COMFORTABLE STAYING WITH? Host 1 Host 2 We look forward to Life is beautiful, so let's hosting you. enjoy it. 2
TRUST ENABLES SOCIAL EXCHANGE (E.G. AIRBNB) 3
A MILLION WAYS THINGS COULD GO WRONG Is this host capable of keeping his/her place clean, safe and comfortable? Does this host care about satisfying my needs during my stay? Will this host stand me up when I show up? Is this host overcharging me? Will I get killed???? 4
DIFFERENT MECHANISMS FOR TRUST ON AIRBNB Reputation System e.g. ratings, reviews Computer Mediated Assurance Policy Communication e.g. customer service, e.g. descriptions, chat, profiles profiles host guarantee 5
A PERSONAL EXAMPLE 6
PREVIOUS WORK ON ONLINE PROFILES Profile as promise. Signaling theory. Ellison, Hancock, Toma (2011) Spence (2002), Donath (2007) 7
RESEARCH QUESTIONS RQ1: What kinds of information do hosts self-disclose to signal their trustworthiness? RQ2: What is the effect of different types of self-disclosure on perceived trustworthiness? RQ3: Do profile-based perceptions of trustworthiness predict choice of host on Airbnb? 8
OUTLINE How does self-disclosure Does perceived Method What do hosts self- Discussion, impact perceived trustworthiness disclose? Limitation, trustworthiness? predict choice? Future Work 9
DEVELOPING TOPICS An iterative, inductive approach Coder 1 5,248 sentences Another 300 300 Sentences from 1,234 profiles Sentences Coder 2 Kappa ranges from Kappa < 0.5 0.5 - 0.8 10
AIRBNB HOST PROFILES DATASET 1.2k 5.2k 8 host profiles sentences topics https://github.com/sTechLab/AirbnbHosts 11
WHAT DO HOSTS DISCLOSE IN PROFILES? 12
WHAT DO HOSTS DISCLOSE IN PROFILES? 13
WHAT DO HOSTS DISCLOSE IN PROFILES? 14
WHAT DO HOSTS DISCLOSE IN PROFILES? 15
WHAT DO HOSTS DISCLOSE IN PROFILES? 16
DIFFERENCES BY HOST TYPE Comparing on-site hosts with remote hosts 80 ON-SITE REMOTE ON-SITE 66 REMOTE 60 56 38% 40 62% 20 0 AVERAGE WORD COUNT * T-tests significant at p<.01 level
DIFFERENCES BY HOST TYPE On-site hosts are more likely to talk about Interests & Tastes and Personality. ON-SITE REMOTE 75% 67% 53% 50% 35% 25% 21% 0% INTERESTS & TASTES PERSONALITY 18 * T-tests significant at p<.001 level
OUTLINE How does self-disclosure Does perceived Method What do hosts self- Discussion, impact perceived trustworthiness disclose? Limitation, trustworthiness? predict choice? Future Work 19
MEASURING TRUST Perceived Trustworthiness Trust Trustworthiness Exhibited by a truster. Attribute of a trustee. What we measure in this work. Potential Guest Potential Guest Host Does Trust Beget Trustworthiness? Kiyonari, Yamagishi, Cook, Cheshire (2006) 20
MEASURING TRUST Trust Game Trust, reciprocity, and social history. Berg (1995) 21
GUEST TRUST SCALE I am confident that the host… Ability Is capable of paying his/her own rent or mortgage. Maintains a clean, safe, and comfortable household. Benevolence Will be concerned about satisfying my needs during the stay. Will go out of his/her way to help me in case of emergency during my stay. Integrity Will stick to his/her word, and be there when I arrive instead of standing me up. Will not intentionally harm, overcharge, or conduct a scam on me. An integrative model of organizational trust. Mayer, Davis, Schoorman (2005) 22
PERCEIVED TRUSTWORTHINESS RATING 1.2k 5 X host profiles judges / profile Host Profile Text Please rate how confident you are about each of the statements per self-introduction based on the text alone. [Maintains a clean, safe, and comfortable household.] 0% 10% 20% …… 80% 90% 100% Not confident Highly confident 23
WHAT DETERMINES PERCEIVED TRUSTWORTHINESS? 24
PERCEIVED TRUSTWORTHINESS BY LENGTH Longer profiles are perceived as more trustworthy — with diminishing returns. “100” means that “I am 100% confident the host is trustworthy”. Ability Perceived Trustworthiness 100 75 50 25 1 10 100 1 25
PERCEIVED TRUSTWORTHINESS BY LENGTH Longer profiles are perceived as more trustworthy — with diminishing returns. Ability Benevolence Integrity Perceived Trustworthiness 100 75 50 25 1 10 100 1 10 100 1 10 100 Word Count 26
WHAT DO HOSTS DISCLOSE IN PROFILES? 27
NUMBER OF TOPICS MENTIONED IN PROFILE 400 336 300 269 COUNT 239 231 200 117 100 0 1 2 3 4 5+ NUMBER OF TOPICS MENTIONED IN THE PROFILE 28
ANALYSIS OF TOPIC COMBINATIONS (1, 2, 3 TOPICS) The Distribution of Perceived Trustworthiness Score for One-topic Profiles (N = 117) Work or Education Host 2 Host 1 Origin or Residence Hospitality Personality Interests & Tastes Travel Life Motto & Values Life is beautiful, We look forward to Relationships so let's enjoy it. hosting you. 29
ANALYSIS OF TOPIC COMBINATIONS (1, 2, 3 TOPICS) The Distribution of Perceived Trustworthiness Score for One-topic Profiles (N = 117) Work or Education Host 2 Host 1 Origin or Residence Hospitality Personality Interests & Tastes Travel Life Motto & Values Life is beautiful, We look forward to Relationships so let's enjoy it. hosting you. 30
ANALYSIS OF TOPIC COMBINATIONS (1, 2, 3 TOPICS) The Distribution of Perceived Trustworthiness Score for One-topic Profiles (N = 117) Work or Education Host 2 Host 1 Origin or Residence Hospitality Personality Interests & Tastes Travel Life Motto & Values Life is beautiful, We look forward to Relationships so let's enjoy it. hosting you. 31
OUTLINE How does self-disclosure Does perceived Method What do hosts self- Discussion, impact perceived trustworthiness disclose? Limitation, trustworthiness? predict choice? Future Work 32
A PERSONAL EXAMPLE 33
MULTIPLE TOPICS + HOSPITABLE LANGUAGE Origin or Residence Work or Education Interests & Tastes Hospitality Travel Relationships Personality 34 Life Motto & Values
BETTER PROMPTS? 35
COMPUTATIONAL APPROACH? A computational approach to politeness with application to social factors. Danescu-Niculescu-Mizil, Sudhof, Jurafsky, Leskovec, Potts (2013) 36
AIRBNB HOST PROFILES DATASET 1.2k 5.2k 8 host profiles sentences topics https://github.com/sTechLab/AirbnbHosts 37
ONE LAST THING (USUALLY) Xiao Ma [1] Jeff Hancock [2] Kenneth Lim Mingjie [3] Mor Naaman [1] [1] Social Technologies Lab, Cornell Tech [2] Department of Communication, Stanford University [3] Department of Computer Science, Cornell University xm75@cornell.edu | maxiao.info 38
BONUS SCENE “Trumping Promises” “You will find staying with me an enriching experience.” https://www.theguardian.com/us-news/video/2015/jun/16/donald-trump-us-president-republicans-video 39
BONUS SCENE https://bit.ly/airbnb-ma 40
ONE LAST THING (REALLY) … e s a e l p � d n A Xiao Ma [1] Jeff Hancock [2] Kenneth Lim Mingjie [3] Mor Naaman [1] https://bit.ly/airbnb-ma [1] Social Technologies Lab, Cornell Tech [2] Department of Communication, Stanford University [3] Department of Computer Science, Cornell University xm75@cornell.edu | maxiao.info 41
Recommend
More recommend
