Information Technology Reference
In-Depth Information
9.4.1 A General Model of Gamification
We suggest a general model of gamification consisting of the following four compo-
nents:
Atask T that needs to be performed.
A set of game design elements g
G
.
A set of users u
U
processing task T enhanced by
G
.
A task-dependent ground truth
f
: U G.
A function class
F
consisting of functions of the form
f
: U G.
The gamification design problem is the problem of selecting a function f
F
that
best approximates the supervisor f .
The ground truth f is a function that assigns each user u a game design element
g that maximizes the expected contribution of u to achieve a prespecified goal. For
users that best perform without any of the game design elements contained in
G
we
include a distinguished symbol
denoting the absence of any design element.
Typically, the ground truth is unknown for most users and therefore needs to be
approximated by a function from some function class
ε
F
based on a small subset
Z = { (
u 1 ,
g 1 ),...(
u n ,
g n ) } ↆ U × G
of training examples. The training set
Z
consists of n users u i with corresponding
design elements g i
for which the ground truth is known.
Note that we do not want to memorize the training examples but rather find
(learn) a function f
=
f
(
u i )
F
that predicts the best fitting design elements for new users
not considered in
Z
.
9.4.2 Learning Problem
There are different ways to select (learn) a function f from
F
in order to approximate
the ground truth f .
One approach describes users u by a feature vector x u . The components of
x u measure different properties of that user such as, for example, click behavior,
mouse movements, and other features. Then a classifier such as the support vector
machine [ 5 ] is trained to learn a model that predicts the best fitting game design
element for new users.
Search WWH ::




Custom Search