Analysis of a Cultural Heritage Game with a Purpose with an Educational Incentive
Analysis of a Cultural Heritage
Game with a Purpose
with an Educational Incentive
Irene Celino, Andrea Fiano and Riccardo Fino
CEFRIEL – Milano, Italy
paper: http://dx.doi.org/10.1007/978-3-319-38791-8_28
Introduction and motivation
Cultural Heritage (CH) Game with a Purpose (GWAP)
Gameplay: quiz to guess a CH asset from four different images
Purpose: ranking images of CH assets
Collateral effect: learning about CH
Evaluation of the CH GWAP:
Effectiveness to achieve its ranking purpose
Ability to engage and retain players
Potential effect of the educational/cultural incentive
Indomilando gameplay
Images of CH architectures in
the city of Milano
Churches, villas, buildings,
schools, towers, …
1 name/title and 4 photos
(of 4 different assets)
Link to official asset record
Map to explore CH assets
the user played with
http://bit.ly/indomilando
Intuitively, a photo is recognizable if it is correctly guessed more frequently
Conservative measure: lower bound of Wilson score interval
But some categories are more recognizable
than others (e.g. buildings are difficult)
correction: standardization
by asset type
Final photo rank score:
Indomilando purpose
Computing photo ranking
p: observed fraction of successes (=photo guessed)
z2
α/2: (1 - α/2) quantile of the standard normal distribution
n: total number of trials (=photo played)
Indomilando purpose
Computing CH asset ranking
Intuitively, an asset rank score corresponds to the average
of its photos’ rank scores
But each asset has a different no. of photos (1-40)
correction for number of photos
But each set of photos can be heterogeneous (mix of good
and bad photos)
correction for photo inhomogeneity
Final asset rank score:
Indomilando purpose
Evaluating ability to rank
Comparison with no. of visits of assets’
Wikipedia pages (when existing)
But no. of Wikipedia visits is more a sign
of “popularity” than “recognisability”
Comparison with ground truth
Best photo of top-10 Wikipedia assets
Manual ranking by 12 users
Rank aggregation weighted by declared
familiarity with Milano CH
Good ability to rank and much more time-effective
Each manual ranker: avg. of 1 min 27 sec to rank 10 assets
Indomilando: 72 players with avg. of 7.5 min to rank ~1400 photos and ~650 assets
Indomilando gaming flavor
Evaluating players engagement
ALP (Average Life Play): 7.5 min
Left-skewed distribution (long tail)
Median < 3 min
Two user groups (+ “outliers”)
Subjective analysis of engagement
through user questionnaire
Players like it
Indomilando cultural flavor
Evaluating educational incentive (1/2)
Possibility to explore CH official records
between game rounds (map exploration)
Two rounds consecutive if < 15 min
Distribution of “exploration” time
Possibility to learn over time to recognize
CH assets
No. of correctly guessed photos as a
function of the played game rounds
Indomilando cultural flavor
Evaluating educational incentive (2/2)
Subjective analysis of cultural incentive through user questionnaire
What motivated you to play? Did you learn anything new about Milano?
guessing it right
leaderboard
learning something new
Conclusions
Indomilando is effective to achieve its ranking purpose
Resulting rank highly correlated to a ground truth
Outcome achieved in a very limited time
Indomilando shows a good engagement potential
Most players find the game fun
User group spending a significantly long time in playing
Indomilando has a learning/educational effect
Players are motivated to acquire new knowledge
Interplay of ranking purpose and educational incentive to be further investigated
Analysis of a Cultural Heritage Game with a
Purpose with an Educational Incentive
Irene Celino, AndreaFiano andRiccardoFino
Thank you!
Additional material can be found at http://swa.cefriel.it/urbangames/indomilando/icwe2016.html
http://bit.ly/indomilando