Sie sind auf Seite 1von 6

A Mobile Phone based reliable interaction tool for

intention-based virtual groups of strangers


- A case of taxi sharing application
Sangwon Choi, Sangyong Gil Zakaria Alkabbab
Computer Science Department, KAIST Telecommunications Department, INSA Lyon
Daejeon, Korea Lyon, France
{sangwonchoi7, i.m.cliet}@gmail.com zakaria.alkabbab@gmail.com

ABSTRACT who, most of the time, know each other or have the same
Communication is a major concern in the global world we interest. Specifically, applications such as Foursquare [5]
are living, especially in sharing activities among strangers and WhosHere [12] implement those services. The present
who speak different languages or have different cultures. work expands these applications into a more profitable field
These differences engender trust issues because we can’t (activity sharing) in a view to make savings. For instance, a
conceive a safe environment where benefits are made by taxi sharing application allows its users to save money by
people who don’t know each other. Giving the possibility to sharing a ride. This system is different from current SNS
users to provide their personal information is one approach applications because users don’t have to meet again. They
to resolve this issue. But, this solution affects people’s meet only for the purpose of making benefits. However,
privacy and safety because of the difficulty to measure several problems arise when strangers have to meet
loyalty and credibility of individuals. sometimes because they are just shy or they don’t feel safe
This area hasn’t been well investigated yet. In fact, the and comfortable with people they don’t know. They
complexity of human behaviors sets a solid barrier facing actually want to interact before meeting strangers.
research works related to collaboration between people Moreover, they don’t like to share their personal
from different communities. Nevertheless, our surveys information so there is a tradeoff between privacy and
attest that communication can build trust and confidence making benefits by sharing activities with strangers. We
among strangers after a reasonable amount of time. In this conducted numerous surveys that led us to consider the
paper, we present the design and implementation of a novel operation time which consists of making homogenous
reliable interaction tool for making real connection among groups and efficiently manages the discussion to reach an
strangers within intention based virtual groups to make agreement quickly. This paper introduces the design and
benefits by sharing activities. implementation of a taxi sharing application which covers
the issues discussed above and emphasizes the usefulness of
Author Keywords
an interaction tool to help users feel safe and comfortable to
collaborate. This application is economical in many aspects
Interaction tool, interaction with strangers, location-based because it allows its users to save money, it is also
social network, privacy, trust, communication environmentally friendly because users share one car so
they save energy and pollute less. We assume that, by
providing simple interaction method, real connection can be
1. INTRODUCTION made without personal information (phone number, sex,
The smartphone market has taken off with the introduction religion…)
of high-tech phones with extensive mobile capabilities,
services like Location Based Service (LBS) and Social This paper is organized as follows. We discuss related
Network Services (SNS) gained popularity within users works in Section 2. In Section 3, we introduce the results of
because they allow dynamic interactions between users the survey we conducted before implementing the taxi
sharing system. Results of the experiment are then analyzed
in Section 4. Followed by a description of the main
Permission to make digital or hard copies of all or part of this work for functions of the application and a graph summarizing its
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
overall architecture as well as its implementation in Section
bear this notice and the full citation on the first page. To copy otherwise, 5. Then, we conduct a small discussion in Section 6 before
or republish, to post on servers or to redistribute to lists, requires prior concluding the paper in Section 8.
specific permission and/or a fee.
2. RELATED WORK To answer this question, we asked “when you take a taxi,
Location-based social networks let users see where their imagine there is somebody with whom you can share taxi.
friends are, search location-tagged content within their Are you going to share the taxi with him or her?”. The
social graph, and meet nearby friends [1]. Based on location results show that 90% (18 people) of participants like to
information, Nan Li and Guanling Chen [1] found four share taxi with strangers. Therefore, we can admit that most
different types of human moving patterns. Moreover, university students want to make groups with strangers if
Foursquare [5] is a representative Location-based social there is an advantage. Though, we cannot be certain that
network application. Similarly, numerous SNS sites merge this will be generalized to all people because all participants
their property with location such as Facebook [4] – Place or in this survey are university students. At least, this survey
twitter [11]. Our work is different because those demonstrates that there are groups who are very motivated
applications are focusing on people who know each other. to meet strangers in order to save money. This experience
On the other hand, WhosHere [12] finds chatting candidates can also give them the opportunity to make new friends.
based on users location information. This is similar to Further, half of the participants want to use the taxi-sharing
handling strangers who are located in close area, but application even if only half of battery capacity is left.
chatting does not require real meeting. In addition, the
situation is not dynamic in this case. By reviewing current
Location-based SNS, we found that it doesn’t address
connections between strangers, so we can consider our Number
work as a contribution to this service. of users
Manually insert the location using map or text 5
Further, Ride sharing and taxi sharing systems have been
developed only on web environment [13] and mobile phone Use from previous destination database 9
environment [9]. In contrast we these works, we are
focusing more on building a safe environment that will Automatically finds the users 6
allow strangers to meet and collaborate.
Table 1: User preference method of making virtual group
3. SURVEYS
To verify issues on making real group from intention-based
virtual group, we applied a taxi sharing system. In the Q2: What kind of methods you want to use to make a
online-taxi-sharing system, users request to share a taxi and virtual group who shares the same interest as you by
find an online group that has same interest (want to share mean of a mobile phone application?
taxi, located at close area, and have same destination). Then, To make real connection from intention-based virtual group
they have to meet in a certain place. among strangers, there must be a virtual group at first. And
To build this system, we have composed 5 interrogations there are lots of ways to make a virtual group which has the
we want to investigate. Therefore, we conducted two off- same intention or interest when using a mobile phone
line questionnaires to understand users’ activity when application. Indeed, the problem is more complicated in
meeting strangers in front of campus cafeteria. mobile devices because people need a simple way to make
a group such a touching or pushing a button once.
The first questionnaire involves Q1 and Q2. It was
completed by 20 people. All participants are KAIST We asked the question “when you enter the destination to
university students (male: 11, female: 9, average age: 20.9) make a taxi sharing group, which method is better for you?”
At first, we thought people would like automatic group
The second questionnaire includes Q3, Q4, and Q5. This recommendation system. But surprisingly, table 1 shows
questionnaire was completed by 45 people (male: 27, that manual insertion is preferred than the automatic way.
female: 18, average age: 22.2). Same as the first Actually, predicting user’s next location is not a simple task.
questionnaire; all participants are KAIST students. Moreover, GPS data is not that accurate to find a specific
area. So, we decided to develop a manually well-defined
Q1: Does it interest you to meet strangers to save graphical interface so that people can use our system just by
money or make benefits by sharing an activity with touching once.
them?
As in social commerce, gathering friends and making a Q3: How far would you go to make a real group?
group to get discounts is common because people do not For a taxi sharing system, there is a distance limitation that
have to meet strangers. However, if they have to make a we use to make groups because group members should
virtual group, wait for other members, then meet face-to- meet in a nearby place. At the same time, people might be
face, and take a taxi with the group members. Thus, our aim ready to wait a longer time or walk a longer distance to
is to know whether people are eager to share a taxi or not meet the other members of the same group if there is more
with strangers. benefit. So we asked two questions; “Imagine you are going
to take a taxi from KAIST to Dunsan (costs about 5,000
KRW). If there are other people with whom you can share with strangers. At the same time, the other members must
the taxi, how far would you walk to meet them?”, and be trustworthy so any kind of interaction will be inevitable
“Imagine you are going to take a taxi from KAIST to to improve the credibility of the system.
Daejeon station (costs about 10,000 KRW). If there are
other people with whom you can share the taxi, how far Q5: How much personal information would you like to
would you walk to meet them?” share?
Through five questions we got the following result; before
meeting strangers, people want to communicate and interact
with them within a virtual group by talking or text-
Duration 5,000 KRW 10,000 KRW messaging. Moreover, they don’t like the idea of sharing
Less than 1 min 7 12 their personal information because it’s private. Therefore
there is a trade-off between privacy and trust. The more
Less than 3 min 8 10 they interact, the better they trust.
Less than 5 min 16 16
INTERACTION TOOL
Less than 10 min 8 3 According to our surveys, we observe that 1) taxi sharing
Less than 20 min 6 4 requires making real groups, but it does not require a lot of
intimacy among group members because they do not have
Table 2: Number of people who will use the taxi sharing to meet again. 2) There must be a minimal credibility to
system according to the amount of money saving/distance become a group member.
of the ride
Based on this observation, we propose an interaction tool
that users can communicate and interact to reach an
The only difference is how much the user can save money.
agreement. When a virtual group is composed, users can
If two people share a taxi, they can just save 2,500 ~ 5000
see where the other members are. Then, members can
KRWs. Although, table 2 shows that most people are
choose place where they want to meet. Those responses will
willing to go to a place five minutes away, this cannot be an
be notified to other members and users can feel safe when
absolute threshold. However, we can infer that people can
they are interacting with each other. Even with this minimal
move to another place to meet strangers.
interaction, we assume that it will be enough to share a taxi
together. Also, position information can be delivered more
correctly by our interaction tool than by voice call or text
How to decide the place to # of people message.
meet?
Public place like taxi stop by 12
system
Mid-place among users by 15
system
Negotiation between users 18
If you are Phone 7
going to
Skype 8
negotiate,
which method SMS 16
is better? Figure 1: interaction tool layout for taxi sharing system,
Messenger 8
location that users want to meet is represented to all group
Table 3: Preference of determining meeting place members when a user touches (chooses) a location

4. EXPERIMENT
To verify that our interaction tool works as we expected, we
Q4: How to decide the place where you have to meet? set the experiment with 3 conditions in table 4.
After virtual group is made, people decide the place where
they will meet. Table 3 shows the preference of determining Problem statement
meeting place. Meeting place can be determined by the In the case of making real-connection from the intention-
system itself or by negotiation. More than half participants based virtual group among strangers in by mean of a mobile
like that the system suggest a meeting place and the rest of phone application, we evaluate the trade-off between users’
them prefer text messaging (SMS and messenger) to voice privacy and trust through our survey results. Likewise, we
message when negotiation is required. That could be mentioned that people don’t want to share their context
explained by the fact that people are reluctant to interact
Implementation details for the experiment environment will
Description be explained in the implementation section.
Case 1 Recommended meeting place by system
Participants
No personal information Our participants are six university students who do not
Case 2 Recommended meeting place by system know each other (average age: 24.8, all male). Table 4
shows the condition of the experimentation. Each subject
Phone number is provided
participated in the task six times and executed tasks with
Our Recommended meeting place by system counter-valancing. Four samples are gathered from the
Interaction Users can dynamically update their experiment for each condition.
Tool location
Table 4: Description of each condition
RESULT
Figure 2 shows the average task completion time of each
condition (our interaction tool: 502.5 sec, condition 1:
easily in a mobile environment. They need some an 496.8 sec, and condition 2: 481.5 sec) with no significant
intuitive structured interface only (5W1H) [6], as well as difference between conditions (t > 0.05). Connection rate
exact and reliable information. gives the same result (75%, 75% and 75%) with a 10
minutes threshold. Because of the small sample size, we
Hypothesis couldn’t see any statistical significance. But the result can
By providing our interaction tool, structured information change by increasing samples and modifying experiment
will be made easily, and people can make real-connection setting.
without exchanging their personal information.
Comments are as follows: “It was good to see the others
Objective suggestion and tell my opinion just clicking”, “I would
We want to show that real connection can be made without always go to the place which is determined by system”,
any personal information using the interaction tool we “Knowing others’ phone number makes me confident”, “I
proposed. like the interaction tool. This is such a fun game”, “I was
late because I had something to do, and the system should
Task be a little bit faster”.
We randomly choose three subjects who do not know each Experiment did show the result that we have expected
other and give them a recommendation meeting place. We because most participants felt that they perform tasks based
set 10 minutes (600 seconds) as a threshold of task success. on the system request in an easy and efficient way.
To do balanced evaluation, we set average distance from However, the situation of experimentation is different from
each subject to less than 5 minutes walking. The reason we that of taxi sharing case. Also, all participants are university
chose three participants is because we can observe the case students so they care less about credibility, and they are not
of two people at the same time. Also, three users are real strangers.
enough to check the multiple user case.
We used two evaluation metrics. The first one is connection
rate which shows the number of times where participants
meet. The other metric is task completion time which
means elapsed time to meet each other. To get correct
completion time, we check the time at the meeting place,
connect participants through the school web page and
explain tasks in advance the day of the experiment. After all
the tasks are over, we take comments from the participants.

Apparatus
In this experiment, we used Android 2.1 platform [1] with
various Android phones (Motorola Motoloy, Samsung
GalaxyS and HTC Desire). For the database, we use a
MySQL server with php programming and C# ADO.NET Figure 2 Task completion time (seconds)
[7] for database execution.
And we used MQTT-based push notification [10] to notify 5. IMPLEMENTATION
each user if his group was made or not. To verify that our taxi sharing system works well, we
implemented a real taxi sharing application in Android
platform. Figure 3 shows the overview of our system. The
application is based on a refined touch-screen graphical geographical coordinates, the number of passengers and the
interface. It allows the user to choose his destination. Then amount of time the user can wait to get a result. The request
a request is generated and sent to a MySQL server. This last handler browses the requests table and finds groups of users
manages the database that contains two tables (requests and whose current and destination coordinates belong to the
groups). The server sends back the results to the application same area. For specifically, the difference between their
in the user’s side. It includes two modules; request handler coordinates doesn’t exceed a certain distance
and group checker. ($MAX_DIST). Further, group members of the group
formed, at the end of each iteration, are notified so that they
can communicate with each other and decide the meeting
place, otherwise the smartphone application suggests a
meeting place (nearest public place) to the members.
Finally, the request handler removes from the database the
requests of the users for whom it found a group. It also
removes expired requests.

User privacy consideration


No personal information is sent to the backend server. The
sensing scripts for our platform capture only hashed
identifiers, and data is secured and anonymized before
aggregate analysis.

Energy efficiency
The GPS sensor used in the taxi sharing application is not
sampled periodically; it’s triggered only when the user
Figure 3 Architecture of taxi sharing system sends a request to the server.

6. DISCUSSION
Message handler
This research is designed to examine trust building among
Message handler is a operating module in the user’s side.
strangers with anonymous exchanges. We define an
When users send a request to the taxi-sharing system, the
anonymous exchange as one in which individuals interact
push-register automatically registers the device at the server
with each and do not know the identity of their partners. We
using MQTT protocol [8]. At the same time, request
disentangle trusting and trustworthiness behaviors using an
manager sends the request to the MySQL server in the
implementation of a novel approach to interaction among
following format:
strangers. Our surveys shows that people are not
– (current GPS data, destination GPS data, time, number of trustworthy when they are uncertain about others’ trusting
people, ttl) intentions. Interestingly, even with a total lack of formal
mechanisms, and little (if any) informal sanctions due to the
Number of people represents the number of people who
minimum scope of social relations, we find evidence of
represent one request and ttl means the maximum waiting
trusting behavior.
time of each user.
7. FUTURE WORK
Group checker
We observe that communication is a marvelous mean that
Group Checker managers group updates, group status and
allows users to build trust before sharing an activity.
notifies group members. It calls request handler periodically.
Therefore more efforts have to made to develop additional
We set cycle duration as one second in our experiment.
features to accelerate the communication process. Future
This value can be decreased, but one second does not affect
works may also address various applications based on the
the efficiency of the system. After that Group checker
same concept introduced in this paper. For instance, we can
retrieves the group table, it calls push notification if there is
conceive an application that finds group members to buy
a change of any record. For example, if there is a group
group tickets for theatre, concerts or soccer game…
with two people at first and one person is added to the
because group ticket prices are often cheaper. Larger scale
group. Then, the group checker notifies again all the
surveys also have to be conducted to better study
members of the group.
interactions among strangers.
Request handler
8. CONCLUSION
This php function runs continuously on the server seeking
In this paper, we discuss the challenges of building trust
groups, it follows a certain process. First, the requests
and finding agreement among strangers who want to make
received by the server are stored in a MySQL database.
benefits by sharing activities. In addition, we had to cope
Each request contains a user ID, current and destination
with trust and privacy issues. We found that an interaction 3. Bhuiyan, T., Xu, Y., Jøsang, A. Integrating Trust with
tool based on communication is the most suitable for this Public Reputation in Location-based Social Networks for
scenario. Then we presented the design, implementation Recommendation Making. International Agent
and evaluation of a novel Mobile Phone based reliable tool Technology Conference, (2008), 107-110.
for intention based virtual groups. Our experiment design 4. Facebook
consists of comparing connection rate between two
different implementations. In the first one, the user provides http://facebook.com
their personal information so that his group members can 5. Foursquare
know about him before meeting him. In the second http://foursquare.com
implementation, the user is free to discuss with the other
members and spontaneously reach an agreement. The 6. Goldthwaite, D., Roberts, W. L. Pragmatic structure in
evaluation results show that the second implementation was appointment-making conversations. Journal of
more adapted to users. Nevertheless, it is important to Psycholinguistic Research, (1993), 22, 579-591.
mention that the surveys are made in South Korea and 7. MySQL connectors in C#
human interactions are very dependent to culture, therefore http://www.mysql.com/products/connector/
results may differ in other communities. Still, we believe
8. MQTT Protocol
our work came up with a major contribution to human
communication area and can serve as a baseline framework http://www-
for more general applications such those discussed in future 01.ibm.com/support/docview.wss?rs=171&uid=swg240
works. 06006
9. Odom, W., Jensen, S., Li, M., Senior travel buddies:
ACKNOWLEDGMENTS sustainable ride-sharing & socialization. CHI 2007, 2079-
We thank Professor Junehwa Song for supporting this 2084
research work and providing helpful comments. We are
10.Push notification for Android
also grateful to You Chongkouk for his collaboration and
idea. http://tokudu.com/2010/how-to-implement-push-
notifications-for-android/
Some of the references cited in this paper are included for
illustrative purposes only. 11. Twitter
http://twitter.com/
REFERENCES
12. WhosHere
1. Li, N., Chen, G. Analysis of a Location-based Social
Network. International Conference on Computational http://myrete.com/
Science and Engineering. (2009), 263-270. 13. Winter, S., Nittel, S. Ad hoc shared-ride trip planning
2. Android SDK by mobile geosensor networks. International Journal of
http://developer.android.com/sdk/android-2.1.html GIS, (2006), 20, 899-916