Game Learning Analytcs for Evidence=Based Serious Games

Summary of a talk by Baltasar Fernandex Manjon at CELDA 2019



ie.ft.com/uber-game

Serious games

  • -         Have been used successfully in many domains – medicine, military
  • -         But low adoption in mainstream education
  • -         So we say we’re working in ‘game-like simulation’

-         Fake news, trolls, e-influencers

http://play.centerforgamescience.com
http://centerforgamescience.org
http://www.re-mission2.org

  • -       Has been formally evaluated

Citizen science

  • -         Uses games for crowdsourcing
  • Also:
    • -         Educational versions of commercial games

Do serious games actually work?

  • -         Very few sg have been formally evaluated
  • -         Evaluation could be as expensive as producing the game  - difficult to get funding
  • -         It is difficult to deploy the game in the classroom

Learning analytics

  • -         Long and Siemens

Game Analytics

  • -         Application of analytics to game dev and research
  • -         Telemetry – info obtained at a distance
  • -         Game metrics – interpretable measures of data related to games
  • -         Mostly used for commercial purposes; proprietary

Business analytics

  • -         From what happened, to why it happened, to what will happen, to how I can make it happen
  • -         Ie., hindsight – insight – foresight
  • -         Needs all the dat
  • -         Now being used in MOOCs, because they have so much data

Game Learning Analytics (GLA)

  • -         Learning analytics applied to serious games
  • -         Collect, analyze and visualize

Uses of GLA

  • -         Game testing – eg., how many finish, avg. time to completion
  • -         Game deployment in class – tools for teachers, eg. ‘stealth’ student evaluation
  • -         Formal game evaluation

RAGE – game analytics (using xAPI)
Beaconing – game deployment

GLA or Informagic?

  • -         Informagic - false expectations of gaining full insight on the game educational experience based on shallow data
  • -         Need to set realistic expectations – most of the games are black boxes

Minimum Requirements for GLA
  • -         Need access to what’s going on during the game
  • -         Need access to the game ‘guts’, or the game must communicate
  • -         Need to understand the meaning of the data – access to developers
  • -         Also must consider ethics of data collection
    • o   Are user informed?
    • o   Is data anonymized
    • o   Note: GDPR – creates an overhead load
GLA structure

  • -         Need to be based on learning objective
  • -         Based on traces + analysis
  • -         Different levels of design – LAM

Experience API

  • -         New defacto standard, becoming an IEEE standard
  • -         e-UCM group in collaboration with ADL for profile for serious games (xAPI-SG)
  • -         xAPI-SG defines a set of verbs, activity types, and extensions
Game trackers / Analytics frameworks as open code
-          http://github.com/e-ucm

Systematization of Analytics Dashboards

  • -         Provided analytics uses xAPI-SG, dashboards do not require additionalconfiguration
  • -         You can also do real-time analytics and warnings – more complex to do
  • -         We were surprised to find how hard it is to make a visualization understandable by the average teacher – eg. Teacher interprets difficulty as ‘you are in Facebook’

uAdventure

  • -         uAdventure tool (on top of Unity)
  • -         game development platform
  • -         includes analytics

Overview of research – 87 papers

  • -         GLA purposes – mostly assessment, n-game behavious; little on interventions\techniques: mostly classical linear analytics, clusters; neural nets not broadly applied
  • -         Stakeholders – teachers came third; not widely deployed
  • -         Focus – to teach, most domains math and science, small sample sizes
  • -         Assessment – mostly pre-post assessments
  • -         Method – 2 steps – game validation phases, game deployment phase

Research questions

  • -         Can we predict student knowledge after playing the game
    • o   With/without pretest
    • o   Can we use for evaluation?
  • -         Need to have greater student numbers for analysis to be useful
  • -         Result – using naïve bayes – yes, we can predict student outcomes
  • -         Not sure about use for evaluation
Case Study

  • -         Game on Madrid Metro used with Down Syndrome students

Case

  • -         Connectado – high school cyberbullying
  • -         Some minigames you can never win
  • -         Result – increase in cyberbullying perception

Simva

  • -         Tool used for scientific validation of serious games
  • -         Goal: to simplify the validation and deployment




Comments

Popular Posts