Can causality accelerate experimentation in software systems?

Andrei Paleyes; Han-Bo Li; Neil D. Lawrence

doi:10.1145/3644815.3644985

edit

Back to publications

Can causality accelerate experimentation in software systems?

Andrei Paleyes, Han-Bo Li, Neil D. Lawrence

Proceedings of the IEEE/ACM 3rd International Conference on AI Engineering, 2024.

Abstract

Software designed using dataflow architecture naturally produces a graphical model of the data transformation process through the system. Interpreting this as a causal graph, we can leverage techniques from causal inference to estimate downstream effects of changes in code components, which can be interpreted as interventions within the causal graph. This allows for less costly software experimentation and can add another layer of protection against undesirable production updates.

Links

Cite this Paper

BibTeX


@InProceedings{can-causality-accelerate-experimentation-in-software-systems,
  title = 	 {Can causality accelerate experimentation in software systems?},
  author = 	 {Paleyes, Andrei and Li, Han-Bo and Lawrence, Neil D.},
  booktitle = 	 {Proceedings of the IEEE/ACM 3rd International Conference on AI Engineering},
  year = 	 {2024},
  doi = 	 {10.1145/3644815.3644985},
  url = 	 {/publications/can-causality-accelerate-experimentation-in-software-systems.html},
  abstract = 	 {Software designed using dataflow architecture naturally produces a graphical model of the data transformation process through the system. Interpreting this as a causal graph, we can leverage techniques from causal inference to estimate downstream effects of changes in code components, which can be interpreted as interventions within the causal graph. This allows for less costly software experimentation and can add another layer of protection against undesirable production updates.
}
}

Endnote

%0 Conference Paper
%T Can causality accelerate experimentation in software systems?
%A Andrei Paleyes
%A Han-Bo Li
%A Neil D. Lawrence
%B Proceedings of the IEEE/ACM 3rd International Conference on AI Engineering
%D 2024	
%F can-causality-accelerate-experimentation-in-software-systems
%R 10.1145/3644815.3644985
%U /publications/can-causality-accelerate-experimentation-in-software-systems.html
%X Software designed using dataflow architecture naturally produces a graphical model of the data transformation process through the system. Interpreting this as a causal graph, we can leverage techniques from causal inference to estimate downstream effects of changes in code components, which can be interpreted as interventions within the causal graph. This allows for less costly software experimentation and can add another layer of protection against undesirable production updates.

RIS


TY  - CPAPER
TI  - Can causality accelerate experimentation in software systems?
AU  - Andrei Paleyes
AU  - Han-Bo Li
AU  - Neil D. Lawrence
BT  - Proceedings of the IEEE/ACM 3rd International Conference on AI Engineering
DA  - 2024/06/11	
ID  - can-causality-accelerate-experimentation-in-software-systems
DO  - 10.1145/3644815.3644985
UR  - /publications/can-causality-accelerate-experimentation-in-software-systems.html
AB  - Software designed using dataflow architecture naturally produces a graphical model of the data transformation process through the system. Interpreting this as a causal graph, we can leverage techniques from causal inference to estimate downstream effects of changes in code components, which can be interpreted as interventions within the causal graph. This allows for less costly software experimentation and can add another layer of protection against undesirable production updates.

ER  -

APA


Paleyes, A., Li, H. & Lawrence, N.D.. (2024). Can causality accelerate experimentation in software systems?. Proceedings of the IEEE/ACM 3rd International Conference on AI Engineering doi:10.1145/3644815.3644985 Available from /publications/can-causality-accelerate-experimentation-in-software-systems.html.