1) The document proposes using gamification as an incentive mechanism for participatory sensing to motivate users and reduce costs for clients.
2) An experiment was conducted with 18 users over one month to obtain a participation probability model and found gamification increased probability while difficulty decreased it.
3) A reward minimization problem is formulated to select users and determine reward points to minimize total cost while ensuring sensing requirements are met.
Gamification-based Incentive Mechanism for Participatory Sensing
1. Gamification-based Incentive
Mechanism for Participatory Sensing
Yoshitaka Ueyama, Morihiko Tamai,
Yutaka Arakawa, Keiichi Yasumoto
Nara Institute of Sciense and Technology
2. Abstract
2014/3/25 crwodsensing 2
Incentive mechanism for participatory sensing
Proposed
Incentive Gamification
Monetary
Incentive
Define & solve a problem for selecting a set of users so
that the cost (rewards) paid by the client is minimized
reducing the cost (rewards) for sensing PoI
Users
(participate in
sensing tasks)
Client
(request sensing tasks)
rewar
d
sensed
info
Who is the
best to ask?
3. Outline
1. Background
2. Related Work
3. Basic Idea
4. Proposed Method
5. Experiment for Deriving Participation Probability
Model
2014/3/25 crwodsensing 3
4. What is participatory sensing?
Collect information from point of interest (PoI) by
asking mobile users to get and send back the
information
2014/3/25 crwodsensing 4
This road is
congested.
This café is not
vacant.
Sunny around
here.
Ex. Application
Road congestion
Vacancy in café/parking
Weather report
5. Challenges in participatory sensing
Need expensive reward for high-burden sensing
2014/3/25 crwodsensing 5
Reward
Pains to input
information
High reward
Client
Too far to PoI!
Embarrassed to
take a photo here
Client want to minimize reward paid to users
e.g., moving, data input, mental load
6. Outline
1. Background
2. Related Work
3. Basic Idea
4. Proposed Method
5. Experiment to Obtain Participation Probability
Model
2014/3/25 crwodsensing 6
7. Related works
(2)Auction-based approach[5]
Reduce rewards by allowing only a user who bids the
lowest price to make a sensing
2014/3/25 crwodsensing 7
Goal is to reduce the cost paid by the client
(1)Game-theory based approach[3]
Use game theory to predict each user’s action and derive the
minimum reward based on the prediction
UsersClient
Reward
[3]L. Jaimes, IEEE INFOCOM 2012 [5] L.Duan, IEEE PerCom 2012
Want to reduce
costs
8. Problem of existing studies
Existing studies focused only on monetary incentive.
2014/3/25 crwodsensing 8
Need other incentive to
increase users’ motivation
Client wants to reduce the cost for sensing PoI
But
Cannot avoid increase of reward for high-burden sensing
Need high reward for increasing users’ motivation
Problem
9. Basic Idea in our study
Employ gamification as incentive mechanism
Motivate users by mental satisfaction
Try to reduce the cost by employing gamification
2014/3/25 crwodsensing 9
proposal
incentive
Gamification
Monetary
incentive
Monetary
incentive Monetary
Incentive
gamificationcost
reduction!
Satisfy users by two incentive
10. Gamification
Incorporate gaming factor into non-gaming activities
Users can enjoy participating activities like game
Gaming factor
Level, title, mission, visualization of ranking among users, etc
Example of services with gamification
foursquare・・・badges depending on the number of check-in
Nike+・・・visualize & share of running-record
2014/3/25 crwodsensing 10
Level, badges,
mission, visualization
Gaming factor
Company activities
Education
SNS
Non-game activities
incorporating
11. Outline
1. Background
2. Related Work
3. Basic Idea
4. Proposed Method
5. Experiment to Obtain Participation Probability
Model
2014/3/25 crwodsensing 11
12. Supposed participatory sensing
2014/3/25 crwodsensing 12
Client
Specifies PoI(Point of Interest) PoI: set of points
Requests sensing of PoI to users nearby
Tell reward points for the users
Users
Determine participation in the requested sensing probabilistically
Receive the specified reward points when completing the sensing
Red circles: PoI
reward
request
PoI: X
0.6
Ask users A and B
for sensing X with
10 pts reward
Prob.
Uses near X
13. Incentive
Monetary incentive
Give reward points to users who completes a sensing
(point can be exchanged to money)
Gamification-based incentive
Level Scheme(≒airlines’ mileage service)
• Level is changed according to holding points
• Users with higher level can get more point than lowers
Ranking Scheme
• Visualize & share ranking among users by their holding points
Badge Scheme
• Users who achieved specified condition get badges and points
e.g., complete sensing 5 times→beginner badge with 10 pts
2014/3/25 crwodsensing 13
level1:
×1.0
level2:
×1.2
14. Reward Points Minimization Problem
2014/3/25 crwodsensing 14
Select a set of users and determining reward
points paid to each user
Objective function: Minimize total reward points
Constraint: probability of each PoI sensed by at
least one user > threshold, e.g., 0.95
Client
Whom should I ask a
sensing?
How much reward
points?
Participate at prob
0.7 when reward is
20pts.
Participate at prob
0.5 when reward is
0.5
15. Greedy algorithm
2014/3/25 crwodsensing 15
CP =
Reward Pts
Prob.
Greedily select a user with high participation
probability and low reward points
select users in descending order of CP
① Select users one by one in descending order of CP until constraint
is satisfied
② If constraint is not satisfied, increase reward pts and repeat step 1
Algorithm flow
Client needs to predict participation prob of each user
16. Outline
1. Background
2. Related Work
3. Basic Idea
4. Proposed Method
5. Experiment to Obtain Participation Probability
Model
2014/3/25 crwodsensing 16
17. Experiment to obtain probability
In order to obtain participation probability,
developed a participatory sensing system based on
foursquare and conducted user study
• Obtain check-in information using FoursquareAPI
• Send e-mail describing a sensing request to the user
• Collect data of participation of users
2014/3/25 crwodsensing 17
Check-in at
restaurant Developed
System
(NAIST Photo2)
Check-in info.
Request(e-mail)
Take a photo of menu of the day
Reward:20 pts.
earned points can be used
to purchase drinks, snacks, etc
(100pts = 100JPY (1USD))
18. Screenshots of the developed system
2014/3/25 crwodsensing 18
Request(e-mail) Ranking Earned badges
Reward= base pts x level-dependent co-efficient
19. Result: probability vs. users
Experiment with/without gamification
18 users, one month, 480 requests
2014/3/25 crwodsensing 19
0
0.2
0.4
0.6
0.8
1
A B C D E F G
Gm有り
Gm無
Participationprobability
Users who received 20 or more requests
Participation probability is increased by gamificationEnthusiasm for gamification is different among users
With Gm
Without Gm
20. 0.4
0.5
0.6
0.7
0.8
0.9
10Pt 12Pt 15Pt 20Pt 30Pt 40Pt
Gm有
Gm無
Result: probability vs. rewards
Relationship between reward pts and probability
2014/3/25 crwodsensing 20
Lv1(10Pt)
Lv2(12Pt)
Lv3(15Pt) Lv5(10Pt×3.0)
Lv4(10Pt×2.0)
Q. Why increase?
A. High level user
have high prob.
Q. Why flat?
A. Reward is too low.
Participationprobability
With Gm
Without Gm
21. Result: probability vs. difficulty
Content of Request
Prob.
with gm without gm
Take photo of landscape 0.93 0.67
Take photo of parking usage 0.89 0.69
Take photo of restaurant’s
limited menu
0.56 0.48
Take photo of congested level
of facility
0.36 0.5
2014/3/25 crwodsensing 21
Easy
Diffi
-cult
Relationship between difficulty and probability
difficulty of request decrease participation probability
22. Observation
Evaluate the effectiveness of gamification
Calculate the number of users required to satisfy the constraint
defined in the minimization problem
2014/3/25 crwodsensing 22
Experimental
Result
With Gm Without Gm
Average of the
Probability
0.71 0.54
The number of users required
constraint With Gm Without Gm
Success Prob.>0.90 2 3
Success Prob.>0.95 3 4
Success Prob.>0.98 4 6
Client can achieve a sensing even though a few users around PoI exist
23. Summary
Propose gamification-based incentive mechanism
for participatory sensing.
Level Scheme, Ranking Scheme, Badge Scheme
Formulate rewards minimization problem
To solve this problem, we conduced an experiment to
obtain the participation probability model
• gamification increase the probability
• reward do not effect the probability
• difficulty of request decrease the probability
2014/3/25 crwodsensing 23
24. Future work
Model an accurate participation probability
based on the experimental result
Evaluate weather gamification can reduce the
total reward points through the simulation study
2014/3/25 crwodsensing 24