Causal Testing Framework

A Causal Inference-Driven Software Testing Framework

The Causal Testing Framework is composed of a :term:causal inference-driven architecture designed for functional black-box testing. It leverages graphical causal inference (CI) techniques to specify and evaluate software behaviour from a black-box perspective. Within this framework, causal directed acyclic graphs (DAGs) are used to represent the expected cause–effect relationships between the inputs and outputs of the system under test, supported by mathematical foundations for designing statistical procedures that enable causal inference. Each causal test case targets the causal effect of a specific intervention on the system under test--that is, a deliberate modification to the input configuration expected to produce a corresponding change in one or more outputs.

Installation

Requirements

• Python 3.10, 3.11, 3.12 and 3.13

To install the latest stable release of the Causal Testing Framework:

pip install causal-testing-framework

or if you want to install with the development packages/tools:

pip install causal-testing-framework[dev]

Alternatively, you can install directly via source:

git clone https://github.com/CITCOM-project/CausalTestingFramework
cd CausalTestingFramework

then to install a specific release:

git fetch --all --tags --prune
git checkout tags/<tag> -b <branch>
pip install . # For core API only
pip install -e . # For editable install, useful for development work

For more information on how to use the Causal Testing Framework, please refer to our documentation.

Note

We recommend you use a 64-bit OS (standard in most modern machines) as we have had reports of the installation crashing on legacy 32-bit Debian systems.

Usage

Note

Example usage can be found in the examples directory.

1. To run the causal testing framework, you need some runtime data from your system, some causal test cases, and a causal DAG that specifies the expected causal relationships between the variables in your runtime data (and any other relevant variables that are not recorded in the data but are known to be relevant).

2. If you do not already have causal test cases, you can convert your causal DAG to causal tests by running the following command.

python -m causal_testing generate --dag_path $PATH_TO_DAG --output_path $PATH_TO_TESTS

3. You can now execute your tests by running the following command.

python -m causal_testing test --dag_path $PATH_TO_DAG --data_paths $PATH_TO_DATA --test_config $PATH_TO_TESTS --output $OUTPUT

The results will be saved for inspection in a JSON file located at $OUTPUT. In the future, we hope to add a visualisation tool to assist with this.

How to Cite

If you use our framework in your work, please cite the following:

This research has used version X.Y.Z (software citation) of the Causal Testing Framework (paper citation).

The paper citation should be the Causal Testing Framework paper, and the software citation should contain the specific Figshare DOI of the version used in your work.

BibTeX Citations

Paper

```
@ARTICLE{Clark_etal_2023,
author = {Clark, Andrew G. and Foster, Michael and Prifling, Benedikt and Walkinshaw, Neil and Hierons, Robert M.
and Schmidt, Volker and Turner, Robert D.},
title = {Testing Causality in Scientific Modelling Software},
year = {2023},
publisher = {Association for Computing Machinery},
url = {https://doi.org/10.1145/3607184},
doi = {10.1145/3607184},
journal = {ACM Trans. Softw. Eng. Methodol.},
month = {jul},
keywords = {Software Testing, Causal Testing, Causal Inference}
}
```

Software (example)

```
@ARTICLE{Wild2023,
author = {Foster, Michael and Clark, Andrew G. and Somers, Richard and Wild, Christopher and Allian, Farhad and Hierons, Robert M. and Wagg, David and Walkinshaw, Neil},
title = {CITCOM Software Release},
year = {2023},
month = {nov},
url = {https://orda.shef.ac.uk/articles/software/CITCOM_Software_Release/24427516},
doi = {10.15131/shef.data.24427516.v1}
}
```

Acknowledgements

The Causal Testing Framework is supported by the UK's Engineering and Physical Sciences Research Council (EPSRC), with the project name CITCOM - "Causal Inference for Testing of Computational Models" under the grant EP/T030526/1.