4. Multiple Paths to Impact
€ Economic
Ý Social
Ħ Policy
á People
! Academic
5. Do
•understand and use the
real problem as a
beacon
•be open to
opportunities
•continually listen
•collaborate
•adapt
•be generous
•approach problems
with integrity
Don’t
•focus only on home runs
•put the research ahead
of the people
•think innovation only
happens in academia
•think scale doesn’t
matter
•think open source is
enough
•think the message
doesn’t matter
6. Do
•understand and use the
real problem as a
beacon
•be open to
opportunities
•continually listen
•collaborate
•adapt
•be generous
•approach problems
with integrity
Don’t
•focus only on home runs
•put the research ahead
of the people
•think innovation only
happens in academia
•think scale doesn’t
matter
•think open source is
enough
•think the message
doesn’t matter
11. Goals
•Lightweight
• Apply to a large system in < 30 minutes
•Flexible
• Multiple kinds of source, relationships between ‘modules’
•Human-centric
• Provide information to human in human’s terms
15. Reflexion Models: Take-aways
•Focus on the problem
• Understanding the implementation
•Notkin guidance
• “Your Ph.D. doesn’t have to be a home run”
• “It is the start of your career and life’s work not the end”
•Openness to opportunity
• Be ready to pitch
• Not the perfect case study but lots of good aspects
Ý Social! Academic
17. Hipikat
• Context:
• The best information about a development is in the heads of the
developers’
• Hypothesis:
• Pertinent artifacts to a task can be recommended from a latent group
memory formed from digital assets recorded as part of the development
• Prototype:
• Built a latent group memory automatically from repository information,
mailing lists, etc.
• Evaluation:
• Wizard-of-oz, controlled lab studies, simulation studies
Çubranić and Murphy, ICSE 2003. Ten year ICSE impact award.
20. Building Hipikat
• Collaboration with IBM when Eclipse was in development
• Likely first academic plug-in built for Eclipse
• IBM personnel participated in lab studies, increasing validity
• Lots of hiccups to overcome
• New development environment and architecture
• Constant upgrading at IBM to be on the latest
21. Hipikat: Take-aways
• Collaboration as an enabler
• Joint collaboration with industry partner
• Scale matters
• Eclipse presented itself as an opportunity we acted upon
• Messages matter
• Hipikat is still first in Google search
• The water cooler description always did and still resonates
• Research can push policy boundaries
• What was ethical to use from open source development?
! Academic á PeopleĦ Policy
23. Who Should Fix This Bug?
• Problem:
• Lots of issues filed in open source
• How to reduce the effort and costs of triaging the bugs
• Proposed solution:
• Use supervised machine learning to learn triage results, like who
should fix a particular bug, which component the bug should be filed
against, etc.
• Evaluation:
• Applied against a few open source systems
• My favourite review comment: “academically honest” because we
explained where it didn’t work
Anvik, Murphy and Hiew, ICSE 2006. Ten-year ICSE impact award.
24. A problem before its time
• We cared about how hard it would be to apply the machine
learning approach to a new system
• Semi-automated configuration of machine learning approach
was important to us for approach to be usable
• At least 1/2 of John Anvik’s thesis was this semi-automated
configuration
• Hard to publish because the problem wasn’t understood (and we
didn’t explain well enough)
• We persevered
25. Who Should Fix This Bug: Take-aways
• Messages matter
• The title captured attention
• Integrity is important
• Outlining context where something works and where it does not is
valuable
• “First base” can be enough
• We opened a new area more than solving a problem completely
• Sometimes impact is not where you expect
• Would I have predicted this as the most-cited paper from the group?
! Academic á People
40. Mylar/Tasktop: Take-aways
•Collaboration
• With Mik Kersten, with open source community, with
Tasktopians, with industry partners, with customers …
•Adaptability
• Pivot to Tasktop Sync
•Open source takes work
• Tasktop supports Mylyn community
!Academic € Economic á People
44. Multiple Paths to Impact
€ Economic
Ý Social
Ħ Policy
á People
! Academic
45. Do
•understand and use
the real problem as a
beacon
•be open to
opportunities
•continually listen
•collaborate
•adapt
•be generous
•be a great academic
Don’t
•focus only on home runs
•put the research ahead
of the people
•think innovation only
happens in academia
•think scale doesn’t
matter
•think open source is
enough
•think the message
doesn’t matter
At the start of the talk…
46. All in terms of “Do”s
•understand and use
the real problem as a
beacon
•be open to
opportunities
•continually listen
•collaborate
•adapt
•be generous
•be a great academic
•focus on the problem at
hand
•see the value in people in
your network
•follow innovations in
industry
•consider whether scale
matters
•realize open source
requires effort
•spend time on the
message
47. Impact can take many forms
Gail C. Murphy
University of British Columbia
Tasktop Technologies Inc. @gail_murphy
Focus on the problem
Be open to possibilities
Think outside the box
Do research with integrity