I will remember that consistency, candor, and compassion should outweigh the performance of a data science algorithm. I will remember that my work may lead to unintended societal consequences, such as inequality, poverty, and other disparities, through algorithmic bias. I will always look for a path to fair treatment and nondiscrimination.

I will evaluate models thoroughly using a comparison to baseline, more than one performance metric, and assessment of per-group performance metrics. I will avoid goal-posting (i.e., chasing improvement in a single metric).

I acknowledge that the practice of Data Science requires more than just technical expertise. I will integrate subject matter experts into the process of my work. I will solicit perspectives on how a given model could have unintended consequences. I will coordinate on plans to mitigate these consequences.

I will aim for my work to be reproducible. To this end, I will thoroughly document all inputs and processes in my work. If possible, I will share my code, results, and testing practices such that others can easily evaluate them.

I recognize that a model needs to be continuously monitored. Quality of data input and model output must be evaluated at regular intervals over the life of the model in order to detect deviations from predetermined standards, combat model drift, and identify unexpectedly harmful predictions.

I will work to provide some channel of recourse for data subjects impacted by a model.