Mathew Hall, Peter Hegedüs, Josip Maras, Marco Mori, Zoltán

Transcription

Mathew Hall, Peter Hegedüs, Josip Maras, Marco Mori, Zoltán
AI Techniques &
Software Engineering
Mathew Hall, Peter Hegedűs, Josip Maras,
Marco Mori, Zoltán Ujhelyi
Leader: Shin Yoo
Mentor: Myra B. Cohen
Who Did We Talk To?
Industry
4
Academia
21
Respondents Know
What AI Techniques in
SE are
Are you aware of uses of AI techniques in software engineering?
No
1
Aware, but not used
9
Aware, and used
15
What do you think of when
you hear the term “AI
Techniques in Software
Engineering”
Are you aware of uses of AI techniques in software engineering?
No
1
Aware, but not used
9
Aware, and used
15
Are you aware of uses of AI techniques in software engineering?
No
1
Aware, but not used
9
No
1
Aware, but not used
9
Why Not?
Never occurred to me
2
Other
4
My problem does not map
4
Why Not?
Never occurred to me
2
Other
4
My problem does not map
4
Why Not?
“I am looking at more directed methods”
“Still haven't had time to learn about it”
“Scepticism about the usefulness and
capabilities.”
“I don't have any reasons to use them”
Never occurred to me
2
Other
4
My problem does not map
4
What Should the
Community Focus On?
15.00
12.00
Abandon
Improve AI techniques
Educate SE community
Widen the research area
Technology transfer
Other
9.00
6.00
3.00
0
# People
Which AI Technique is
Most Relevant in SE?
Which class of AI techniques do you see as most relevant to Software Engineering tasks?
Planning
1
NLP
7
Machine learning
10
Search
7
What Stage Can
Benefit the Most?
Design & architecture
3
Implementation
1
Maintenance
6
Other
3
Requirements engineering
1
Testing & debugging
10
Summary
• People are aware of the use of AI
techniques in SE
• 95% of respondents consider AI in SE to be
emerging or evergreen
• Only 13% of people expect AI techniques
to automatically solve problems
Lessons Learned
• Integration of AI techniques and human
engineers should be seamless
• Participants thought more of machine
learning than SBSE
• Some participants thought AI techniques
are oversold and not sound
• Divergent ideas about maturity
Threats to Validity
• Some questions dependent on previous
answers
• 25 participants
• Questionnaire design
Final Comment
“This survey gets interrupted
by people n”