Causal testing: understanding the root causes of defects
summary
Claim: automated causal experiments take existing failing tests, use fuzzing and automated test generation to produce minimally-different passing and failing executions that reveal why a failure occurs. Built: a proof-of-concept tool that generates and compares inputs and execution paths. Found: participants report the similar passing tests point to root causes. Other uses: causal fairness testing for bias and targeted testing of ML components by flipping single attributes and checking outcome changes.
AI-written summary · about
chapters (5)
- 0:00Start
- 4:17Causal Testing
- 5:15What Causal Testing Does
- 6:30What Else Can Causal Testing Do
- 8:21Testing Machine Learning Based Software
related talks
38mTesting with two failure seeking missiles: fuzzing and property based testing
43mCan you Test it
19mCAFault: Enhance Fault Injection Technique in Practical Distributed Systems via...
26mAutomated testing what is the point
30mFuzzing Testing: Practical Applications
59m.NET Testing Techniques You Didn’t Know You Needed
16mTransforming Test Automation with AI and Machine Learning
44mHow AI is Exposing the Existential Crisis of Software Testing