The Boulder Police Department has a number of ways in which officers create police report narratives, one of them is Draft One. Draft One is a software as a solution that uses body worn camera data to automatically draft a police report narrative for an incident. Safeguards ensure that every report is reviewed and approved by a human officer, ensuring accuracy and ownership of the information.

For the most in-depth information related to Draft One, visit their website at www.axon.com/products/draft-one.

Yes, Draft One narratives can contain mistakes, but the technology is continually being improved.

To mitigate the risk of reports being submitted that contain errors, Draft One emphasizes the importance of personnel reviewing drafts generated by Draft One and requires users to sign off on the accuracy of a draft before they are able to use it in a report.

The quality of the narrative generated by Draft One is often tied to the quality of the audio captured by the body camera. Audio quality and the resulting Draft One narrative can be of lower quality in the following situations:

  • Chaotic incidents
  • Incidents where multiple people are speaking at the same time
  • Incidents where unsupported languages are being spoken
  • Serious crimes
  • Extended interviews

Users have reported that even if the draft generated by Draft One isn’t ideal, it is often a good starting point. The draft lets users review what happened, helps overcome “writers block” by giving them a starting point, and can provide details they may have forgotten.

Axon ran a double-blind randomized control trial study in which 24 experts reviewed 120 narratives. Some of the narratives were generated using Draft One and some were not. The experts included District Attorneys, Command Staff, and equity and inclusion scholars. The experts were asked to score each narrative for five different dimensions: Comprehensiveness, Neutrality, Objectivity, Terminology, and Coherence.

Scores for Comprehensiveness, Neutrality, and Objectivity were the same for narratives generated with the help of Draft One compared to narratives created without Draft One. Draft One narratives performed significantly better where Terminology was concerned. Additionally, Draft One narratives contained significantly less incriminatory language (e.g., hit, assaulted, fled) compared to narratives that were created without using Draft One.

Data stored in Draft One, Axon, is secure.