Successfully reported this slideshow.
Your SlideShare is downloading. ×

[devcom2019] Behavioral Economics for Game: Cases of User Data Analysis

Loading in …3
×

Check these out next

1 of 56 Ad
1 of 56 Ad

[devcom2019] Behavioral Economics for Game: Cases of User Data Analysis

Download to read offline

This session will introduce three cases of applying behavioral economics to analyze game users with a massive amount of data. Sentience's Co-founder and Research Lead, Hyeyon Kwon, will describe meta-model of player's motivations, which reflects four basic human needs of rewards - personal satisfaction, personal rewards, social interaction, public recognition - and elaborate 1) how to quantify and parameterize each motivation from data, 2) build econometric models, and 3) analyze the data to draw implications.

This session will introduce three cases of applying behavioral economics to analyze game users with a massive amount of data. Sentience's Co-founder and Research Lead, Hyeyon Kwon, will describe meta-model of player's motivations, which reflects four basic human needs of rewards - personal satisfaction, personal rewards, social interaction, public recognition - and elaborate 1) how to quantify and parameterize each motivation from data, 2) build econometric models, and 3) analyze the data to draw implications.

Advertisement
Advertisement

More Related Content

Advertisement

[devcom2019] Behavioral Economics for Game: Cases of User Data Analysis

  1. 1. Behavioral Economics for Game: Cases of User Data Analysis
  2. 2. Who we are • Behavioral economics-based AI company • Analyze games and user’s behavior and their motivation- to-play • Develop and service recommendation system based on the user’s motivation-to-play
  3. 3. What I am going to talk about • Introduce meta model for user’s motivation-to-play • Introduce how to measure each motivation based on player’s log data • Introduce analysis cases
  4. 4. Meta model for motivation-to-play SocialPersonal Intrinsic Personal Satisfaction Social Interaction Personal Rewards Public Recognition Extrinsic • McGuinness M. 2015. Motivation for Creative People: How to Stay Creative While Gaining Money • Rigby, S. and Skinner, T. GDC 2014, The Importance of Player Autonomy: Motivating Sustained Engagement through Volition and Choice • Baron J. GDC 1999, Glory and Shame: Powerful Psychology in Multiplayer Online Games • Ryan, R.M. and Deci, E.L.2000. Intrinsic and Extrinsic Motivations: Classic Definitions and New Directions
  5. 5. Meta model for motivation-to-play SocialPersonal Intrinsic Personal Satisfaction Social Interaction Personal Rewards Public Recognition Inherently interesting or enjoyable Extrinsic • McGuinness M. 2015. Motivation for Creative People: How to Stay Creative While Gaining Money • Rigby, S. and Skinner, T. GDC 2014, The Importance of Player Autonomy: Motivating Sustained Engagement through Volition and Choice • Baron J. GDC 1999, Glory and Shame: Powerful Psychology in Multiplayer Online Games • Ryan, R.M. and Deci, E.L.2000. Intrinsic and Extrinsic Motivations: Classic Definitions and New Directions
  6. 6. Meta model for motivation-to-play SocialPersonal Intrinsic Extrinsic Personal Satisfaction Social Interaction Personal Rewards Public Recognition Inherently interesting or enjoyable Leads to a separable outcome • McGuinness M. 2015. Motivation for Creative People: How to Stay Creative While Gaining Money • Rigby, S. and Skinner, T. GDC 2014, The Importance of Player Autonomy: Motivating Sustained Engagement through Volition and Choice • Baron J. GDC 1999, Glory and Shame: Powerful Psychology in Multiplayer Online Games • Ryan, R.M. and Deci, E.L.2000. Intrinsic and Extrinsic Motivations: Classic Definitions and New Directions
  7. 7. Meta model for motivation-to-play Intrinsic Extrinsic • McGuinness M. 2015. Motivation for Creative People: How to Stay Creative While Gaining Money • Rigby, S. and Skinner, T. GDC 2014, The Importance of Player Autonomy: Motivating Sustained Engagement through Volition and Choice • Baron J. GDC 1999, Glory and Shame: Powerful Psychology in Multiplayer Online Games • Ryan, R.M. and Deci, E.L.2000. Intrinsic and Extrinsic Motivations: Classic Definitions and New Directions Personal Satisfaction Social Interaction Personal Rewards Public Recognition Inherently interesting or enjoyable Leads to a separable outcome SocialPersonal Need for self expression
  8. 8. Meta model for motivation-to-play Intrinsic Extrinsic • McGuinness M. 2015. Motivation for Creative People: How to Stay Creative While Gaining Money • Rigby, S. and Skinner, T. GDC 2014, The Importance of Player Autonomy: Motivating Sustained Engagement through Volition and Choice • Baron J. GDC 1999, Glory and Shame: Powerful Psychology in Multiplayer Online Games • Ryan, R.M. and Deci, E.L.2000. Intrinsic and Extrinsic Motivations: Classic Definitions and New Directions Personal Satisfaction Social Interaction Personal Rewards Public Recognition Inherently interesting or enjoyable Leads to a separable outcome SocialPersonal Need for self expression Other people’s opinions matter
  9. 9. Who are social-intrinsic people? SocialPersonal Intrinsic Extrinsic • Harmony • Collaboration • Loyalty • … Personal Satisfaction Social Interaction Personal Rewards Public Recognition
  10. 10. Who are social-extrinsic people? Personal Satisfaction Social Interaction Personal Rewards Public Recognition SocialPersonal Intrinsic Extrinsic • Recognition • Appreciation • Awards and Prizes • …
  11. 11. Who are personal-extrinsic people? Personal Satisfaction Social Interaction Personal Rewards Public Recognition SocialPersonal Intrinsic Extrinsic • Money • Privileges • Opportunities • …
  12. 12. Who are personal-intrinsic people? Personal Satisfaction Social Interaction Personal Rewards Public Recognition SocialPersonal Intrinsic Extrinsic • Pleasure • Learning • Meaning • …
  13. 13. 3 Steps for implementing motivation model into games RECOMMENDDEFINE FIND (1) Define and parametrize the motivation you want to find Personal Satisfaction Social Interaction Personal Rewards Public Recognition
  14. 14. 3 Steps for implementing motivation model into games RECOMMENDDEFINE FIND (2) Find each player’s motivation based on the player’s log data Install Friend’s introduction The game is not what player expected Tutorial Curiosity No interest Quick character selection Character selection Complex game Interest in certain characters Induce billing Item curiosity Character interest No interest Fast level- up 다른 캐릭터 경험 PvP IAP Set goals Recommend item that player is not interested Login record Usage of gold record Acquiring gold record Purchase record Play info Character choice record
  15. 15. 3 Steps for implementing motivation model into games RECOMMENDDEFINE FIND (3) Send recommendation message based on the motivation
  16. 16. The result of automizing the three steps: Retention rate increased by 13% User engagement increased by 45% In-app purchase increased by 8% Real case example: Result of applying motivation-based recommendation system in mobile game
  17. 17. Analysis methodology How to define & parametrize & analyze the motivation model? Behavioral Economics & Econometrics • Behavioral economics: Study of the effect of psychological, emotional, social factors on economic decisions of individuals • Econometrics: Statistical methods for empirical analysis BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE
  18. 18. Case 1: Play Garden Social interaction in the social game (Play Garden) Personal Satisfaction Social Interaction Personal Rewards Public Recognition
  19. 19. Case 1: Play Garden Social interaction in the social game (Play Garden) Research Question: How shutdown law changed the social interactions among players? Shutdown law in South Korea forbidding children under the age of 16 to play online video games between 00:00~06:00 Personal Satisfaction Social Interaction Personal Rewards Public Recognition
  20. 20. Case 1: Play Garden Social interaction in the social game (Play Garden) Social interaction Network effect Find social interaction in Play Garden As more users use a service, they create more values, and more people join Personal Satisfaction Social Interaction Personal Rewards Public Recognition 1 BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE
  21. 21. Personal Satisfaction Social Interaction Personal Rewards Public Recognition Case 1: Play Garden Social interaction in the social game (Play Garden) More users ↔ the number of concurrent users Use a service & create more value ↔ play time Social interaction Network effect As more users use a service, they create more values, and more people join Find social interaction in Play Garden1 BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE
  22. 22. Case 1: Play Garden Social interaction in the social game (Play Garden) Parametrize social interaction The number of concurrent users and their play time Social interaction Network effect Personal Satisfaction Social Interaction Personal Rewards Public Recognition • The number of concurrent users • Total play time for each user for each day • Other variables that may affect the players: • Level for each user on specific day • Gold each user has on specific day 2 BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE
  23. 23. Case 1: Play Garden Social interaction in the social game (Play Garden) Personal Satisfaction Social Interaction Personal Rewards Public Recognition The number of concurrent users The number of concurrent users The user’s play time - The user’s level - Gold - Weekend - School vacations - Demographics Build a model measuring social interaction and the effect of shutdown policy 3 BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Source: Kwon and Suh (2013)
  24. 24. Case 1: Play Garden Social interaction in the social game (Play Garden) Build a model measuring social interaction and the effect of shutdown policy Personal Satisfaction Social Interaction Personal Rewards Public Recognition 3 The number of concurrent users The number of concurrent users The user’s play time - The user’s level - Gold - Weekend - School vacations - Demographics Shutdown Law BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Source: Kwon and Suh (2013)
  25. 25. Case 1: Play Garden Social interaction in the social game (Play Garden) BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE 2SLS A methodology for analyzing the effect of independent variables on the dependent variable when there are feedback loops in the model Select the right methodology and analyze 4 Chow Test A methodology for testing the presence of a structural break at a specific event in time series data The number of concurrent users The number of concurrent users The user’s play time - The user’s level - Gold - Weekend - School vacations - Demographics Shutdown Law Source: Kwon and Suh (2013)
  26. 26. Case 1: Play Garden Social interaction in the social game (Play Garden) BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Select the right methodology and analyze 4 Personal Satisfaction Social Interaction Personal Rewards Public Recognition Source: Kwon and Suh (2013) The number of concurrent users The number of concurrent users The user’s play time - The user’s level - Gold - Weekend - School vacations - Demographics Result: Before shutdown law No network effect
  27. 27. Case 1: Play Garden Social interaction in the social game (Play Garden) BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Select the right methodology and analyze 4 Personal Satisfaction Social Interaction Personal Rewards Public Recognition Source: Kwon and Suh (2013) Result: After shutdown law Network Effect & Reverse Network Effect The number of concurrent users The number of concurrent users The user’s play time - The user’s level - Gold - Weekend - School vacations - Demographics Shutdown Law
  28. 28. Case 1: Play Garden Social interaction in the social game (Play Garden) BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Select the right methodology and analyze 4 Personal Satisfaction Social Interaction Personal Rewards Public Recognition Source: Kwon and Suh (2013) Result: After shutdown law Network Effect & Reverse Network Effect More concurrent users Play time Fewer concurrent users
  29. 29. Case 1: Play Garden Social interaction in the social game (Play Garden) BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Select the right methodology and analyze 4 Personal Satisfaction Social Interaction Personal Rewards Public Recognition Source: Kwon and Suh (2013) Result: After shutdown law Network Effect & Reverse Network Effect Play time Reverse network effect More concurrent users Fewer concurrent users
  30. 30. Case 1: Play Garden Social interaction in the social game (Play Garden) BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Select the right methodology and analyze 4 Personal Satisfaction Social Interaction Personal Rewards Public Recognition Source: Kwon and Suh (2013) Result: After shutdown law Network Effect & Reverse Network Effect Play time Network effect More concurrent users Fewer concurrent users
  31. 31. Case 2: Ever Planet – Social distance Public recognition in MMORPG (Ever Planet) Personal Satisfaction Social Interaction Personal Rewards Public Recognition
  32. 32. Case 2: Ever Planet – Social distance Public recognition in MMORPG (Ever Planet) Research Question: Does social distance influence the player’s behavior in games? Social distance is the gap between different groups; social class, race, gender, or sexuality. In the virtual world, social distance is the level difference between existing users and new users. Personal Satisfaction Social Interaction Personal Rewards Public Recognition
  33. 33. Case 2: Ever Planet – Social distance Public recognition in MMORPG (Ever Planet) Public recognition Social distance Find public recognition in Ever Planet Personal Satisfaction Social Interaction Personal Rewards Public Recognition 1 BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE The gap of levels among users; mostly between existing users and new users. Source: Kim and Lee (2013)
  34. 34. Case 2: Ever Planet – Social distance Public recognition in MMORPG (Ever Planet) Parametrize public recognition the difference between the user's level and the average level of all users. Public recognition Social distance Personal Satisfaction Social Interaction Personal Rewards Public Recognition 2 BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE • 𝑃𝑖𝑗: Total played time for player j during day i • 𝑆𝐷𝑖𝑗 = 𝐿𝑖,𝑗 − 𝐴𝑣𝑔(𝐿𝑖) Social distance: difference between player j’s level during day i and average level of all users during day i • 𝐴𝑃𝑖𝑗 = 𝑘=1 𝑖−1 𝑃𝑖,𝑗: Accumulated played time of player j from day 1 to day i-1 • 𝑁𝑖: Number of players played in day i Source: Kim and Lee (2013)
  35. 35. Case 2: Ever Planet – Social distance Public recognition in MMORPG (Ever Planet) Personal Satisfaction Social Interaction Personal Rewards Public Recognition Build a model measuring public recognition and its effect on the player’s behavior 3 BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Source: Kim and Lee (2013) Social distance The number of concurrent users The user’s play time - Accumulated playtime - Number of days passed - Weekend - School vacations
  36. 36. Case 2: Ever Planet – Social distance Public recognition in MMORPG (Ever Planet) BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Fixed effect A methodology for analyzing panel data when group means are fixed for each group. Select the right methodology and analyze 4 Source: Kim and Lee (2013) Social distance The number of concurrent users The user’s play time - Accumulated playtime - Number of days passed - Weekend - School vacations
  37. 37. Case 2: Ever Planet – Social distance Public recognition in MMORPG (Ever Planet) BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Select the right methodology and analyze 4 Personal Satisfaction Social Interaction Personal Rewards Public Recognition Source: Kim and Lee (2013) Result: As users have been playing more, they play the less today Social distance The number of concurrent users The user’s play time - Accumulated playtime - Number of days passed - Weekend - School vacations (-)
  38. 38. Case 2: Ever Planet – Social distance Public recognition in MMORPG (Ever Planet) BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Select the right methodology and analyze 4 Personal Satisfaction Social Interaction Personal Rewards Public Recognition Source: Kim and Lee (2013) Result: But the social distance makes people behave differently; the higher level, the longer they play Social distance The number of concurrent users The user’s play time - Accumulated playtime - Number of days passed - Weekend - School vacations (+)
  39. 39. Case 2: Ever Planet – Social distance Public recognition in MMORPG (Ever Planet) Personal Satisfaction Social Interaction Personal Rewards Public Recognition Source: Kim and Lee (2013) How to manage the social distance? Agent-Based Model A computational model for simulating the actions and interactions of autonomous agents The simulation results show that when the player-matching system is applied, the game retention is significantly increased. The example of Divisions and Tiers systems in online / offline sports Agent-Based Model simulation when the player- matching system is applied in the game to reduce social distance Players under Lv.50 Players under Lv.30 No player- matching # of players time
  40. 40. Case 2: Ever Planet – Game characteristics Personal rewards in MMORPG (Ever Planet) Research Question: How to categorize people with personal rewards among various people? In MMORPG, there are different types of people, and we want to distinguish people with different kinds of motivations. Personal Satisfaction Social Interaction Personal Rewards Public Recognition
  41. 41. Case 2: Ever Planet – Game characteristics Personal rewards in MMORPG (Ever Planet) Find the characteristics of the game1 BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Personal Satisfaction Social Interaction Personal Rewards Public Recognition Dungeons Playing parties
  42. 42. Case 2: Ever Planet – Game characteristics Personal rewards in MMORPG (Ever Planet) Dungeons Playing parties Personal Satisfaction Social Interaction Personal Rewards Public Recognition Parametrize the characteristics of game2 BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE • 𝑃𝑇𝑖𝑗: Total played time for player j on the day i • 𝑃𝑇𝑖−1,𝑗: Total played time for player j on the day i-1 • 𝐴𝑉𝐺_𝑃𝑆𝑖,𝑗: Average party size the player j joined on the day i • 𝑆𝑇𝐷_𝑃𝑆𝑖,𝑗: Standard deviation of party size the player j joined on the day i • 𝐸𝑖,𝑗: The number of dungeons the player j entered on the day i Playing types; the number of dungeons entered and the number of playing parties of each user
  43. 43. Case 2: Ever Planet – Game characteristics Personal rewards in MMORPG (Ever Planet) Build a model measuring the characteristics of the game 3 BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE AVG_PartySize The user’s play time (t-1) The user’s play time (t) - The number of dungeon entries - STD_PartySize (how different parties the user joined) Personal Satisfaction Social Interaction Personal Rewards Public Recognition
  44. 44. Case 2: Ever Planet – Game characteristics Personal rewards in MMORPG (Ever Planet) BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Random effect A methodology for analyzing panel data when there are individual effects. Select the right methodology and analyze 4 AVG_PartySize The user’s play time (t-1) The user’s play time (t) - The number of dungeon entries - STD_PartySize (how different parties the user joined)
  45. 45. Case 2: Ever Planet – Game characteristics Personal rewards in MMORPG (Ever Planet) BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Select the right methodology and analyze 4Result: All variables in the model affect the user's play time significantly. AVG_PartySize The user’s play time (t-1) The user’s play time (t) - The number of dungeon entries - STD_PartySize (how different parties the user joined) Personal Satisfaction Social Interaction Personal Rewards Public Recognition
  46. 46. Case 2: Ever Planet – Game characteristics Personal rewards in MMORPG (Ever Planet) BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Select the right methodology and analyze 4Result: We can find the players with personal rewards by removing the effect of party size and the interaction effect of party size and the user's play time. AVG_PartySize The user’s play time (t-1) The user’s play time (t) - The number of dungeon entries - STD_PartySize (how different parties the user joined) Personal Satisfaction Social Interaction Personal Rewards Public Recognition
  47. 47. Case 3: Mobile Personal Health Record Personal Satisfaction Social Interaction Personal Rewards Public Recognition Personal satisfaction in mPHR (My Chart in My Hand)
  48. 48. Case 3: Mobile Personal Health Record Personal satisfaction in mPHR (My Chart in My Hand) Research Question: Do people with personal satisfaction have higher retention rates? Health management function has a self-monitoring feature, which represents a personal intrinsic personality. Personal Satisfaction Social Interaction Personal Rewards Public Recognition
  49. 49. Case 3: Mobile Personal Health Record Personal satisfaction in mPHR (My Chart in My Hand) Personal satisfaction Self-monitoring Find personal satisfaction in mPHR1 BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Tracking and updating the user’s health information Personal Satisfaction Social Interaction Personal Rewards Public Recognition
  50. 50. Personal Satisfaction Social Interaction Personal Rewards Public Recognition Case 3: Mobile Personal Health Record Personal satisfaction in mPHR (My Chart in My Hand) Parametrize personal satisfaction The average and the frequency of usage 2 BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Source: Lee et al. (2018) Personal satisfaction Self-monitoring • 𝐴𝑉𝐺_𝑆𝑀𝑖: Average usage of self-monitoring function by patient i • 𝑆𝑇𝐷_𝑆𝑀𝑖: Standard deviation of usage of self-monitoring function by patient i
  51. 51. Case 3: Mobile Personal Health Record Personal satisfaction in mPHR (My Chart in My Hand) Build a model measuring each feature’s effect on user retention 3 BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Source: Lee et al. (2018) Personal Satisfaction Social Interaction Personal Rewards Public Recognition The average usage of self- monitoring function The standard deviation of self- monitoring function usage User retention - Demographics - The level of illness) The average usage of other functions The standard deviation of other function usages
  52. 52. Case 3: Mobile Personal Health Record Personal satisfaction in mPHR (My Chart in My Hand) BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Cox proportional hazard A methodology for analyzing the expected time until an event of interest happens Select the right methodology and analyze 4 Source: Lee et al. (2018) The average usage of self- monitoring function The standard deviation of self- monitoring function usage User retention - Demographics - The level of illness) The average usage of other functions The standard deviation of other function usages
  53. 53. Case 3: Mobile Personal Health Record Personal satisfaction in mPHR (My Chart in My Hand) BUILD A MODELQUANTIFY PARAMETRIZE ANALYZE Select the right methodology and analyze 4 Source: Lee et al. (2018) Result: People with personal satisfaction have higher retention rates. Personal Satisfaction Social Interaction Personal Rewards Public Recognition The average usage of self- monitoring function The standard deviation of self- monitoring function usage User retention - Demographics - The level of illness The average usage of other functions The standard deviation of other function usages (+) (-)
  54. 54. Wrap up • Define and find the motivation of players in games • Personal Satisfaction • Social Interaction • Personal Rewards • Public Recognition • Find the players with analysis methodology • Recommend items and contents based on the motivation
  55. 55. Reference • McGuinness M. 2015. Motivation for Creative People: How to Stay Creative While Gaining Money • Rigby, S. and Skinner, T. GDC 2014, The Importance of Player Autonomy: Motivating Sustained Engagement through Volition and Choice • Baron J. GDC 1999, Glory and Shame: Powerful Psychology in Multiplayer Online Games • Ryan, R.M. and Deci, E.L.2000. Intrinsic and Extrinsic Motivations: Classic Definitions and New Directions, Contemporary Educational Psychology 25, 54–67 • Kwon, H., Suh, C. 2013, The Effect of the Shutdown System on Social Games, Business Administration Research 6, 1-12 • Kim, M., Lee, B. 2013, Are There Too Many Superheroes? Analysis of the Social Distance in Massive Multiplayer Online Role Playing Game, https://ssrn.com/abstract=2330090 • Lee K, Kwon H, Lee B, Lee G, Lee JH, Park YR, et al. (2018) Effect of self- monitoring on long-term patient engagement with mobile health applications. PLoS ONE 13(7): e0201166. https:// doi.org/10.1371/journal.pone.0201166
  56. 56. Thank you. Homepage: www.sentience.rocks Hyeyon Kwon: h.kwon@sentience.rocks Booth: C4

×