Crowdsourcing Software Evaluation

Authors:
Nada Sherief, Nan Jiang, Mahmood Hosseini, Keith Phalp, Raian Ali
{nsherief, njiang, mhosseini, kphalp, rali} @bournemouth.ac.uk
Presented by:
Nada Sherief

Agenda
 Introduction
 Crowdsourcing Evaluation
 Related Work
 Method
 Results
 Research Challenges
 Conclusion
 Acknowledgements
13 May 2014 EASE'14 | London, UK 2

Introduction(1/2)
 Crowdsourcing is a new form of problem solving, which is typically
online and relies on a large number of people.
 Crowdsourcing is a flexible business model and it requires a less strict
recruitment and contracting process. This does not mean that the
process should be open without limits.
 Crowdsourcing is typically used for non-critical tasks and tasks which
naturally require input from the general public and where the right
answer is based on people’s acceptance and dynamics.

Introduction(2/2)
 The stakeholders of software evaluation include users who utilize the
software to reach their needs and expectations, i.e. requirements.
 Users’ acceptance and efficient use of the software is a main subject of
evaluation.
 In a dynamic world, users’ acceptance needs to be captured throughout
the life time of the software to stay up-to-date to communicate their
perception of the role and quality of software at runtime.
 The ultimate goal of this proposal is to maximize the efficiency of
software evaluation and its scalability to cater for complex and highly
variable systems where the power of the crowd, when systematically
coupled, could become a great aid.

Agenda
 Introduction
 Related Work
 Method
 Results
 Conclusion

Crowdsourcing Evaluation(1/2)
 Traditional evaluation methods, which is mainly a design-time activity,
have a range of limitations including the following:
 Limitations in predicting and simulating the actual context of use.
 The unstructured and varied ways in which users provide their
feedback typically in a natural language.
 Capturing the opinion of only an elite group of users.

Crowdsourcing Evaluation(2/2)
 Crowdsourced evaluation has various potential benefits in comparison
to centralized approaches:
 An ability to evaluate software while users are using it in practice
(i.e. at runtime).
 The access to a large crowd enables fast and scalable evaluation.
 An ability to maintain the evaluation knowledge up-to-date
 An access to a wider and diverse set of users and contexts of use that
are unpredictable by designers at design time
 Enabling users to introduce new quality attributes and
requirements.

Agenda
 Introduction
 Related Work
 Method
 Results
 Conclusion

Related Work(1/3)
 There are several established approaches where the role of users is
central, such as: User centered design, User Experience, Agile
methodology, and Usability Testing.
 Recently, more work has been directed towards inventing systematic
methods for representing and obtaining users’ feedback.
 Processes for continuous and context-aware user input were
proposed.
 An empirical study on users’ involvement for the purpose of
software evaluation and evolution has been conducted.
 The crowd feedback was also advocated for shaping software
adaptation.

Related Work(2/3)
 In general, when designing an empirical study in software engineering,
engaging the necessary type of participants and appropriate number is
always a challenge.
 Researchers are often required to perform trade-offs to be able to
perform the study.
 The crowdsourcing paradigm and the mTurk platform have been used
in evaluating the usability of a school’s website.
 The advantages are claimed to be more participants’ involvement, low
cost, high speed, and various users’ backgrounds.
 The disadvantages include lower quality feedback, less interactions,
more spammers, less focused user groups.

Related Work(3/3)
 Most of the existing studies in crowdsourcing in general, and those
exploiting the paradigm for software evaluation, advocate the use of
the paradigm and use commercial platforms, such as mTurk
(https://www.mturk.com/).
 The literature is still limited in providing engineering approaches and
foundations to develop crowdsourcing platforms for software
evaluation.
 Currently, the design and conduct of feedback acquisition are heavily
reliant on developers’ creativity.
 We still need to investigate and devise systematic approaches when
designing feedback requests and aid developers with proper tools.

Agenda
 Introduction
 Related Work
 Method
 Results
 Conclusion

 We took an empirical approach by conducting a multi-session focus
group study, which is a popular technique of qualitative research in
software engineering .
 Both junior and senior software developers were invited to join the first
session where the emphasis of this session was to:
 understand how software developers normally gather user feedback
 how they think a good feedback should be structured
 how they collaborate and communicate with users in the
development as this could inform the way we design feedback
requests
Method(1/3)
12 May 2014 EASE'14| London, UK 13

 The second session was conducted with regular software users who are
used to provide feedback.
 The emphasis of this session was to:
 explore the ways that users would like feedback requests to look like
 what drives them to provide feedback
 their concerns for not getting involved enough and also for being
involved more than what they expect
 This session was also used to investigate their motivations to take
part in projects and learn their experience from that participation
Method(2/3)

Method(3/3)
 A total of 15 volunteers, 8 males and 7 females aged between 18 and 40,
were invited to participate in the focus group study. There were 8
participants in the first session and 7 participants in the second
session.
 These participants mainly came from Egypt and UK with various
backgrounds ranging from management, student, research and IT and
had different experiences in using software and providing feedback.
 Most participants were already familiar with the notion of
crowdsourcing and they have used it in the past for simple tasks such
as collecting the notes for lectures, using programming forums to get
solutions for certain coding and debugging problems.

Agenda
 Introduction
 Related Work
 Method
 Results
 Conclusion

Results(1/3)

Thematic Area Theme Example
Subject Method “Snapshots, Text, or Audio”
Clarity “reach a clear problem specification”
Specificity “specific to the software’s features”
Structure Timing “real-time feedback”
Level of Detail “giving detailed feedbacks”
Measurement “a group of predefined keywords”
“structured in a specific way”
Specificity “give feedback to specific problems”
Results(2/3)

Results(3/3)
Thematic Area Theme Example
Engagement Recognition “a friendly confirmation for participation”
“more personalized options for feedback”
Value “it encourages users to give feedback if they can meet with
analysts to discuss problems in some ways”.
Channel “the interactions should be very simple”
Transparency “it would increase the users’ trust and willingness to give
feedback if they know the cycle of how their feedback will be
used”
Involvement Privacy “would like to stay anonymous”
“it is important if the user can control who is able to see his
feedback”.
Response “the software’s speed of response to my feedback affects my
willingness to give feedback”
Support “there can be videos to explain to the users what they can do (in
order to provide feedback)”
Rewards “the system should do some favor to the user who gave useful
feedback that helped enhance the system”
“trying new features or versions for free if the user gave good
feedback”
“users who gave useful feedback can be accredited to the use to
increase their reputations”

Agenda
 Introduction
 Related Work
 Method
 Results
 Conclusion

Research Challenges (1/2)
 Translation of evaluation metrics and users judgment to terms and
language which are perceivable by users.
 Metrics to decide the right crowd in terms of engagement and expertise
in the requirements and software functionalities to evaluate.
 Ensuring privacy when the evaluation is established as a collaborative
activity
 Aggregation of highly diverse feedback coming from large scale crowd
with different interests and expertise.

Research Challenges (2/2)
 Balancing between user-friendliness of interaction methods and
precision of collected evaluation feedback.
 Balancing between reward mechanisms, which should incentivize the
crowd, and quality of provided feedback.
 Capture of the context of use in which the software was used and
evaluated to increase feedback expressiveness.

Agenda
 Introduction
 Related Work
 Method
 Results
 Conclusion

Conclusion
 This paper proposed a systematic development of crowdsourcing-based
solution to software evaluation.
 While the concept of crowdsourcing is shown to be promising
considering the increasing complexity and diversity of contexts for
current systems, there is still a lack of foundations on how to engineer
it and ensure correctness and maximize quality.
 This paper focused on the activity of interacting with users and getting
their feedback on software quality as one important step for a holistic
approach of crowdsourcing software evaluation.

Agenda
 Introduction
 Related Work
 Method
 Results
 Conclusion
 Acknowledgments

Acknowledgements
 The research was supported by an FP7 Marie Curie CIG grant (the
SOCIAD Project) and by Bournemouth University – Fusion Investment
Fund (the BBB, BUUU and VolaComp projects) and the Graduate
School Santander Grant for PGR Development.

Crowdsourcing Software Evaluation

More Related Content

What's hot (13)

Similar to Crowdsourcing Software Evaluation (20)

More from Engineering and Social Informatics (ESOTICS) (17)

Recently uploaded (20)

Crowdsourcing Software Evaluation