Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

To share or not to share

A @SMLabTO Research Talk By @ANNE_SUPHAN

  • Login to see the comments

  • Be the first to like this

To share or not to share

  1. 1. TO SHARE OR NOT TO SHARE: PRIVACY PROTECTION AND BOUNDARY MANAGEMENT WHEN “#SHARENTING” ON INSTAGRAM Dr. Anne Suphan (University of Hohenheim) @anne_suphan eMail: anne.suphan@uni-hohenheim.de Dr. Anatoliy Gruzd(Ryerson University) @gruzd eMail: gruzd@ryerson.ca
  2. 2. DISCLAIMER To preserve the anonymity of the users, the images presented were modified using the ‘curly hair’ filter in the Prisma application.
  3. 3. „Sharenting is [the] shorthand term denoting when parents share information about themselves and their children online. - Blum-Ross, A., & Livingstone, S. (2017).
  4. 4. Self-presentation as parents on Instagram: Why do parents (over)share photos of children & parenthood on Instagram?
  5. 5. BENEFITS OF SHARENTING ➤ Social Participation: „I would like family and friends to participate in my life.“ ➤ Pride Confirmation: „I am proud of my child and therefore I would like to show it to the whole world.“ ➤ Envy of others: „It feels good to show my child to others and maybe some even get a little bit envious.“ ➤ Perceived Convenience: „In the case of an extensive network of friends and family, it is of course advantageous if one can show his children without immense effort.“ (Wagner, A., & Gasche, L. A. (2018))
  6. 6. Self-presentation of „parenting“ on Instagram: What is needed to be identified as parents on Instagram and what does it costs?
  7. 7. COSTS FOR THE CHILD ➤ Privacy risks for the child: The risk to lose the control over private information about the child. „The photo is quickly distributed and might fall into wrong hands.“ ➤ Face risks for the child: The risk that the child being embarrassed, being the victim of cyber-bullying or simply being negatively evaluated. „There is a fundamental risk that the child will be ashamed of the picture later.“ (Wagner, A., & Gasche, L. A. (2018))
  8. 8. DIGITAL DILEMMA OF PARENTING SELF-PRESENTATION ➤ Presentation of one’s own identity as a parent means making public aspects of a (potentially vulnerable) child’s life. PRIVACY PROTECTION ➤ At the same time, because they are the parent, they are precisely the person primarily responsible for protecting child’s privacy. (Blum-Ross, A., & Livingstone, S. (2017)) VS.
  9. 9. ➤ Actively cover the child’s face with emoticons or to take a picture without showing a face (40%*). ➤ Share pictures with harmless content (37%*). ➤ Reduce the number of posts by sharing only special moments (5%*) ➤ Hide information like name, location and age (2%*). ➤ Share pictures only as long as the child is under 2 years old (1%*). STRATEGIES TO DEAL WITH THE DILEMMA * These are self-reported privacy intentions/ strategies of parents! (Wagner, A., & Gasche, L. A. (2018))
  10. 10. PRIVACY PARADOX ATTITUDES ➤ Internet users’ attitudes show that users are highly concerned about their privacy and the collection and use of their personal information. BEHAVIOUR ➤ Individuals reveal personal information for relatively small rewards (for drawing the attention of peers). (Kokolakis, S. (2017)) VS.
  11. 11. RESEARCH QUESTIONS How do parents deal with the digital dilemma of boundary management („digital self vs. digital child“)? To what extent are privacy protection practices influenced by parents general digital presentation of lifestyle (and self)? 1 2
  12. 12. THEORY … lifestyle and digital presentation of self.
  13. 13. „ - Stebbins, R. A. (1997). A lifestyle is a distinctive set of shared patterns of tangible behaviour that is organised around a set of coherent interests or social conditions or both, that is explained and justified by a set of related values, attitudes, and orientations and that, under certain conditions, becomes the basis for a separate, common social identity for its participants.
  14. 14. DETERMINANTS OF LIFESTYLE LIFESTYLE age gender ethnic origin spatial context social background
  15. 15. THE (IN)VISIBLE habitus digital presentation of self digital presentation of lifestyle „Black Box“ (invisible/ incorporated) Visible privacy attitudes privacy behaviour
  16. 16. DATA COLLECTION … finding a promising approach.
  17. 17. Approach Idea Pro Con #motherhood compare with #fatherhood, identify gender as an influence regional /cultural unspecific #motherhoodun plugged „realistic“ representation of motherhood as representation of lifestyle regional cultural unspecific, gendered #familylife „realistic“ representation of family life as representation of lifestyle gender neutral regional/ cultural unspecific #mamablogger compare with #papablogger, identify gender as an influence cultural specific define themselves as „blogger“ #lebenmitkind (life with a child) „realistic“ representation of family life as representation of lifestyle cultural specific, genderneutral, generalist COMPARISON OF DIFFERENT APPROACHES
  18. 18. APPROACH #LEBENMITKIND(ERN) Netlytic. Instagram data #lebenmitkind (German language) ** 08/01/2018-08/07/2018 Final** dataset. N=2994 postings of public accounts
  19. 19. DATA CLEANING Final dataset. N=2994 postings Profile information. Collecting profile information on the n=200 accounts (most frequently posting accounts). ★ Posts ★ Followers ★ Following ★ Profile description ★ Demographics
  20. 20. Profile information. Collecting profile information on the n=200 most frequently posting accounts. ★ Posts ★ Followers ★ Friends ★ Profile description ★ Demographics DATA CLEANING Online-Shops (n=34). Deleted for further analysis. Unavailable (n=1). Deleted for further analysis.
  21. 21. DATA CLEANING Celebreties (n=24). (more than 5000 followers) Micro-Celebreties (n=46). (1000-5000 followers) Non-Celebreties (n=95). (less than 1000 followers) Profile information. Collecting profile information on the n=200 most frequently posting accounts. ★ Posts ★ Followers ★ Friends ★ Profile description ★ Demographics
  22. 22. DATA CLEANING Profile information. Collecting profile information on the n=165 most frequently posting accounts. ★ Posts ★ Followers ★ Friends ★ Profile description ★ Demographics Final test dataset. N=862 images & descriptions of 162 Instagram accounts
  23. 23. DATA ANALYSIS … dealing with images.
  24. 24. ANALYSING 862 IMAGES… … manually or automated? Both!
  25. 25. DATA ANALYSIS STRATEGY - RQ1 Automated image analysis to identify relevant images that show children Manual image analysis to analyse privacy protection within the images How do parents deal with the digital dilemma of boundary management? Quantitative analysis
  26. 26. CLARIFAI VS. GOOGLE’S CLOUD VISION Using clarifai.com API. Automated analysis of 757 images. ➤ More detailed categories. ➤ Not only analysing images objects. Also describing them. Using Cloud Vision API. Automated analysis of 757 images. ➤ Less detailed categories but lower probability threshold. ➤ Mainly analysing objects at images.
  27. 27. CLARIFAI - EXAMPLE A indoors child family people girl baby 95 % 96 % 98 % 98 % 99 % 99 %
  28. 28. CLOUD VISION - EXAMPLE A shoulder joint leg girl standing arm 72 % 74 % 75 % 80 % 83 % 87 %
  29. 29. CLARIFAI - EXAMPLE B people recreation fun adult leisure vehicle child 93 % 94 % 95 % 95 % 97 % 98 % 99 %
  30. 30. CLOUD VISION - EXAMPLE B water vehicle fun leisure tourism vacation recreation 68 % 71 % 71 % 72 % 72 % 73 % 92 %
  31. 31. CLARIFAI VS. GOOGLE’S CLOUD VISION Using clarifai.com API. Automated analysis of 757 images. ➤ More detailed categories. ➤ Not only analysing images objects. Also describing them. Using Cloud Vision API. Automated analysis of 757 images. ➤ Less detailed categories but lower probability threshold. ➤ Mainly analysing objects at images. Using the clarifai API.
  32. 32. CODING PRIVACY PROTECTION* A. Privacy protection through face covering (FC) 1= approx. 100% of face is complete visible 2= more than 50% of face is visible 3= less than 50% of face is visible 4= 0% of face is visible (e.g. only back part of head, only body is on image) 5= filters are used to cover largely parts of face 👶 👶 👶 😎 👶
  33. 33. CODING PRIVACY PROTECTION* A. Degree of nudity (N) 0= body is completely covered or not visible 1= child wears underwear or swimwear (legs and arms are naked) 2= child wears a diaper (legs, arms and upper body are naked) 3= child is completely naked (whole body is naked) 👚👖 👙
  34. 34. CODING PRIVACY PROTECTION - EXAMPLES FC=4, N=3 FC=3, N=1 FC=5, N=0
  35. 35. CODING PRIVACY PROTECTION Coding was applied manually to all images that were classified by Clarifai as (n=601). ➤ child OR ➤ girl OR ➤ boy OR ➤ baby OR ➤ family OR ➤ person. n=145 didn’t showed any children but parents. n= 456 images were coded concerning privacy protection.
  36. 36. FIRST RESULTS … of the pre-study.
  37. 37. N min max average S.D. age of account owner 260 22 42 28,83 5,277 # of children 549 1 4 1,35 0,606 likes/ post 755 0 2769 90,29 179,034 DESCRIPTIVES - WHO IS SHARENTING ON INSTAGRAM?
  38. 38. N % single parent 11 1 % married 340 45 % missing 407 54 % Total 758 100 % DESCRIPTIVES - WHO IS SHARENTING ON INSTAGRAM? Could this be interpreted as an indicator for differences in lifestyle? N % female 691 91 % male 26 3 % both parents 33 4 % missing 8 1 % Total 758 100 % In line with previous research: „[M]others take on the responsibility of sharing content about their children more than fathers do.“ (Ammari et al., 2015)
  39. 39. DESCRIPTIVES - FACE PROTECTION N % cumulative % 1 114 25 25 2 66 15 40 3 63 14 53 4 182 40 93 5 31 7 100 Total 456 100 On 25% of shared images the face of child/ children are completely visible. 👶 👶 👶 😎 👶
  40. 40. DESCRIPTIVES - FACE PROTECTION N % cumulative % 1 114 25 25 2 66 15 40 3 63 14 53 4 182 40 93 5 31 7 100 Total 456 100 On the other side, 40% of shared images do not show the face of child/ children. 👶 👶 👶 😎 👶
  41. 41. DESCRIPTIVES - DEGREE OF NUDITY n % cumulative % 0 354 78 78 1 54 12 90 2 45 10 99 3 3 1 100 Total 456 100 On most shared images with a child (78%) the body of the child is covered with clothes. 👚👖 👙
  42. 42. DISTRIBUTION OF FACE PROTECTION BY NUDITY On 31% of images that show a child almost naked, the face of the child is completely visible. 0 1 2 3 Total 1 26 % 15 % 31 % 0 % 25 % 2 14 % 15 % 20 % 0 % 15 % 3 13 % 28 % 4 % 33 % 14 % 4 40 % 37 % 38 % 67 % 40 % 5 7 % 6 % 7 % 0 % 7 % Total 100 % 100 % 100 % 100 % 100 % 👶 👶 👶 😎 👶 👚 👖 👙
  43. 43. DISTRIBUTION OF FACE PROTECTION BY NUDITY On 40% of images that show a child completely covered by clothes, the face of the child is not visible at all. 0 1 2 3 Total 1 26 % 15 % 31 % 0 % 25 % 2 14 % 15 % 20 % 0 % 15 % 3 13 % 28 % 4 % 33 % 14 % 4 40 % 37 % 38 % 67 % 40 % 5 7 % 6 % 7 % 0 % 7 % Total 100 % 100 % 100 % 100 % 100 % 👶 👶 👶 😎 👶 👚 👖 👙
  44. 44. Spearman- Rho followers following posts coefficient ,371** ,417** sign. 0,000 0,000 N 755 755 THE MORE YOU GIVE, THE MORE YOU GET! Spearman- Rho following likes/ post followers coefficient ,596** ,627** sign. 0,000 0,000 N 755 755
  45. 45. THE MORE YOU GIVE, THE LESS YOU CARE? ☞The older the parents and the more often they post parenting content on Instagram, the less they protect the face of their children. Spearman-Rho age frequency face protection coefficient -,226** -,113* sign. 0,004 0,016 N 159 455
  46. 46. THE MORE NAKED, THE MORE POPULAR? ☞The degree of nudity correlates positive with the amount of follower and likes of a post. Spearman-Rho follower likes/ posts degree of nudity coefficient ,101* ,111* sign. 0,032 0,018 N 455 455
  47. 47. WHAT’S NEXT? … work in progress.
  48. 48. DATA ANALYSIS STRATEGY - RQ2 (WIP) Qualitative ethnography of n=10 instagram accounts Automated image analysis of all images posted by the accounts To what extent are privacy protection practices influenced by parents general digital presentation of lifestyle (and self)? Mixed methods: Quantitative and qualitative analysis
  49. 49. ETHNOGRAPHY Selection criteria. < 1.000 followers (youngest) child is between 0-3 years old btw. 50-500 posts
  50. 50. ETHNOGRAPHY Subsample description. n=10 10 female 7 married, 1 single parent ∅ 181 posts (min. 53; max. 416) ∅ 289 follower (min. 86; max. 742) ∅ 24,7 years old (min. 22; max. 28) ∅ 0,46 posts per day (min. 0,09; max. 10,2) ∅ 2,5 face protection (min. 1; max. 5) ∅ 0,7 degree of nudity (min. 0; max. 2)
  51. 51. 3 TYPES OF SHARENTING PROFILES I. It’s all about me! II. It’s all about my child and my family! III. Me, my life and my family!
  52. 52. ANNA: IT’S ALL ABOUT ME ★ 22 years old, married ★ 1 son (6-12 months old) ★ 54 posts ★ instagram account since July 11 2018
  53. 53. ANNA: IT’S ALL ABOUT ME ★ Most images are selfies and #ootd, ★ Anna’s body and face is mostly complete visible, ★ Brand and price of the outfits are named.
  54. 54. ANNA: IT’S ALL ABOUT ME ★ Images representing „mothering“ can be described as situations of giving love, hugs and kisses. ★ Anna is always very present in these images. ★ There is only 1 image showing the child with its father.
  55. 55. ANNA: IT’S ALL ABOUT ME ★ In the beginning the child was represented with >50% face protection. ★ Now more and more images were shared without any face protection.
  56. 56. ANNA: IT’S ALL ABOUT ME woman fashion portrait 32 % 44 % 65 % 65% of all images on Anna’s Instagram account were categorised by clarifai to show a woman.
  57. 57. CLARA: IT’S ALL ABOUT THE CHILD ➤ 28 years old, married ➤ 1 son (14 months old) ➤ 53 posts ➤ instagram account since September 25th 2017
  58. 58. ★ This is the only image showing Clara! CLARA: IT’S ALL ABOUT THE CHILD
  59. 59. CLARA: IT’S ALL ABOUT THE CHILD ★ Most pictures show everyday life situation with a child. ★ In the beginning images of the child had 100% face protection. ★ Now approx. 50% of the face is visible.
  60. 60. CLARA: IT’S ALL ABOUT THE CHILD ★ The father is often represented as involved in everyday life activities.
  61. 61. child family indoors 23 % 30 % 54 % 54% of all images on Clara’s Instagram account were categorised by clarifai to show a child. CLARA: IT’S ALL ABOUT THE CHILD
  62. 62. child woman indoors 22 % 34 % 34 % 34% of all images on Jessi’s Instagram account were categorised by clarifai to show a child or a woman. JESSI: ME, MY LIFE AND MY FAMILY!
  63. 63. FIRST QUANTITATIVE RESULTS ANALYSING LIFESTYLE ➤ The more often images the shows „situations of playing“, „family pictures“, „pictures of children“ or „pictures of toys“ are on the account, the lower is the minimal value of face protection on this account. The more child-orientated the represented lifestyle is, the less parents care about privacy protection on the account. ➤ The more often the account represents images of furniture or interior or several rooms, the lower is the maximum degree of nudity one the account. The less child-orientated the represented lifestyle is, the more parents care about privacy protection on the account.
  64. 64. A SURPRISING FINDING … that needs further research.
  65. 65. AS A SOCIOLOGIST, GENDER DISCRIMINATION IS ALWAYS AN ISSUE… Pearson girl boy cute coefficient ,375** 0,006 sign. 0,000 0,860 N 757 757 beautiful coefficient ,408** -,140** sign. 0,000 0,000 N 757 757 pretty coefficient ,287** -,120** sign. 0,000 0,001 N 757 757 fashion coefficient ,248** -,120** sign. 0,000 0,001 N 757 757 What is the reason for these discriminating results? Is it in the picture or in the algorithm?
  66. 66. boy: 99,8% girl: 98,5% cute: 0,0% beautiful: 0,0% GENDER BIAS IN THE ALGORITHM? boy: 0% girl: 99,4% cute: 94,7% beautiful: 98,6%
  67. 67. FOUR MESSAGES TO TAKE HOME … 1. Although parents know privacy protection strategies, 50% share pictures with their children's sensitive visual information. 2. The privacy protection behaviour depends on the demographics as well as the usage behaviour of the parents. 3. Images that contain a lot of sensitive visual information of children are more popular among users. 4. If the Instagram account is a representation of a lifestyle that focuses primarily on children, sensitive visual information about children are more likely to be shared.
  68. 68. ACKNOWLEDGEMENT ➤ This work was supported in part by the Ryerson University’s Social Media Lab Visiting Scholar Program and the Ontario/ Baden-Württemberg Faculty Research Exchange Program. ➤ Special thanks to Oluwatomilayo Adegbite for his help and support with the automated image analysis.
  69. 69. THANK YOU FOR YOUR ATTENTION!

×