Bug reporting forms
don’t talk back.
End users don’t know what developers need.
Templates give little guidance — and no feedback when a step is missing or wrong.
Introducing
BURT
A chatbot for interactive bug reporting.
B
BURTReporting a bug in Mileage
What went wrong? Describe the problem you saw.
the app crashed
NO MATCH
I couldn’t match that to the app. Could you describe it in more detail?
The average fuel economy shows a NaN value
Is this the screen with the problem?
Yes
Did you perform any of these steps?
Parse every sentence.
BURT parses each message into the parts that matter: who did what, to which part of the app.
“The average fuel economy shows a NaN value”
subjectaverage fuel economy
actionshows
objectNaN value
Check it against the app.
Descriptions are matched to screens in a model of the app, built from automated exploration and real usage traces.
fuel economy · NaN value
APP EXECUTION MODEL
Suggest the next steps.
BURT ranks likely paths to the problem screen, then shows the steps as screenshots to pick from.
REPORTED STEPS
1Launch the appdefault
2Tap “Add fill-up”suggested
3Type “5” in Gallonssuggested
4Save the car fill-upsuggested
Evaluated in two user studies with
BURT reports vs. reports from a template-based bug tracker
12 of 18reporters found BURT easy or somewhat easy to use
A study participant, on BURT’s screenshots
“I liked the screenshots a lot, very easy to report the process to reproduce a bug.”
From static forms
to interactive bug reporting.
Toward Interactive Bug Reporting
for (Android App) End Users
Yang Song · Junayed Mahmud · Ying Zhou · Oscar Chaparro · Kevin Moran · Andrian Marcus · Denys Poshyvanyk
ESEC/FSE 2022
ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering