Jason Jefferies: Yiting, tell us about yourself and your role at Dunlap.

Yiting Liu: I’m Yiting Liu, a Research Associate I at Dunlap and a Ph.D. student in Human Factors and Applied Cognition at NC State University. My work focuses on human factors and applied research, and I enjoy using research methods and data to address practical, real-world questions.

At Dunlap, I work across multiple applied research projects and support different stages of the research process, including survey development, data collection, data analysis, and reporting. My academic training in human factors and statistics has given me experience with both experimental and quantitative methods, which I draw on depending on the needs of each project.

JJ: What is the project called that we are talking about today?

YL: Develop and Test Drug-Positive Driver Detection Cues.

JJ: Who are our partners on this project?

YL: Dunlap has worked with many partners throughout the course of this project. During the time I have been involved, the main partners I have seen collaborating on the project have been law enforcement agencies from different jurisdictions across the country. These agencies provided our team with a large number of DUI case reports involving cannabis, along with corresponding toxicology reports and other relevant case information.

Their contributions have been essential to the project. The case reports and supporting toxicology information provided the foundation for the dataset that our team has been able to organize, review, code, and analyze. Having access to real-world cases from different agencies also allows us to examine patterns across a broad collection of cannabis-related DUI cases. We are very appreciative of the time and effort these agencies and their teams put into gathering and sharing these materials, as their support made the data collection and subsequent analysis possible.

JJ: Who else at Dunlap Research is working on this project with you?

YL: I’ve had the valuable opportunity to work with the entire Dunlap 352 team throughout this project. For much of the project, I worked closely with Alan Mintz (Research Associate) on coding the arrest reports and continuously refining our coding schema as we learned more from the data. During this process, Mikey Pritchard (Systems Analyst) also provided a great deal of support, particularly through his technical guidance and suggestions as we worked through different coding and data-related challenges.

I’ve also had the opportunity to work with you (Research & Communications Associate Jason Jefferies) while organizing and refining the coding schema, including providing data reformatting and other supporting work to help prepare the data for the next stages of the project. Although I did not work directly with Dylan Hewitt (Research Associate), he played an important role earlier in the project by managing and organizing the toxicology data we received. That work provided an important foundation for much of the data integration and analysis that followed.

Throughout the project, I’ve worked with our two PIs, Dennis Thomas (President) and Adam Smith (Principal Research Associate), and learned a great deal from their experience and guidance, especially in how to approach the analysis, management, and presentation of complex research data.

Overall, one of the most valuable parts of this project for me has been the opportunity to work with and learn from the entire team. Everyone has brought different expertise and perspectives to the project, and that collaboration has been an important part of moving the work forward.

JJ: Tell us what the project is about and what it is trying to accomplish.

YL: The Develop and Test Drug-Positive Driver Detection Cues project focuses on identifying cues that may help law enforcement officers recognize drivers who are potentially impaired by marijuana. While there is already a well-established body of research and guidance around detecting alcohol-impaired driving, identifying marijuana impairment presents a different and more complex challenge.

Our work looks at real-world DUI cases, including arrest reports and toxicology information, to identify and examine driving behaviors, physical signs, and other observations that may be associated with marijuana-impaired driving. The goal is to better understand which cues, or combinations of cues, may be useful and reliable indicators for law enforcement. Ultimately, the project is working toward developing an evidence-based set of cues that can better support officers in identifying potential marijuana-impaired drivers in the field.

JJ: What is your role on the project?

YL: My role on the project has evolved as the project has progressed. I initially worked with the team to code arrest reports using our coding schema, while also helping refine and expand the schema as we identified new cues in the reports. I also supported the extraction of additional information from the arrest reports, such as incident times, blood draw times, case types, and other details needed for the analysis.

As the project moved into the later stages, a major part of my role shifted toward data management and analysis. I worked on restructuring and integrating information from multiple datasets to make sure the different pieces of case, toxicology, and coded cue information could be connected and used together. From there, I conducted preliminary analyses and developed visualizations to support team discussions and help inform decisions about the direction and next steps of the analysis.

More recently, I’ve also worked with the team to prepare the data for external expert review. This has involved taking fairly complex underlying datasets and developing more organized, user-friendly review materials so that our collaborating experts can efficiently review the cues, coding decisions, and supporting information and provide their feedback.

JJ: Tell us about coding. What does the coding process for this project involve?

YL: For this project, “coding” mainly involves systematically reviewing the narrative information in arrest reports and turning relevant observations into structured data that can be analyzed. We review each report to identify relevant driving behaviors and other observations documented by the officer before the administration of the Standardized Field Sobriety Tests (SFSTs).

When we identify a relevant observation, we assign it to a corresponding cue based on the project’s coding schema. Because real-world arrest reports can vary considerably in how officers describe similar observations, part of the process also involves determining how different descriptions should be consistently categorized. When we encounter observations that are not adequately captured by the existing schema, we discuss them as a team and refine or expand the coding schema as needed.

The overall goal of the coding process is to systematically convert information from a large collection of narrative arrest reports into consistent, structured data that can then be used for subsequent analysis.

JJ: What sort of documents are you producing for this project, and what purpose do they serve?

YL: As mentioned, the ultimate goal of the project is to take the large amount of information contained across many arrest reports and other supporting data and eventually translate it into a much more concise and practical set of materials that can be tested and ultimately used by law enforcement agencies.

With that goal in mind, the materials I have provided to the team serve several different purposes. Much of my recent work has focused on developing integrated datasets that bring together information from different sources, particularly toxicology results, coded cues, and other case-level information. These datasets provide the foundation for the team’s ongoing analysis and help us move from the original reports toward the specific cues and findings that are most relevant to the project.

I’ve also supported the development of user-friendly review materials for the external experts working with our team. These materials organize the underlying data in a way that allows the experts to efficiently review our cue classifications, definitions, and supporting information and provide additional recommendations based on their expertise. Each of these materials represents a different step toward turning a large and complex collection of real-world data into something that can eventually be efficiently evaluated and applied in practice.

JJ: What is the most surprising thing you have learned while working on this project?

YL: One of the most surprising things for me has been seeing how many different types of dangerous driving behaviors officers have documented in these marijuana-related DUI cases. After reading hundreds of real arrest reports, I really get a sense of how impaired driving can affect people on the road and the kinds of risks it can create for both the driver and others around them.

Seeing these real cases has also made the value of this project much more apparent to me. It reminds me that the work is not just about analyzing data; we are ultimately trying to provide law enforcement with better tools to identify potentially impaired drivers and help make the roads safer.

JJ: How does this project compare to other projects you are working on?

YL: This project has been different from other projects I’ve worked on in a couple of ways. First, it was my first opportunity at Dunlap to work on a project that involves close collaboration with law enforcement agencies. It has been really interesting to see how research can connect directly with the work law enforcement officers are doing in the field.

It was also my first opportunity at Dunlap to work with qualitative data on this scale. Working with hundreds of narrative arrest reports has been a unique experience, especially because the project involves taking detailed, real-world information and systematically organizing it into data that can support further analysis.

JJ: Is there anything we haven’t discussed about this project that you would like to mention?

YL: One thing I would add is that impaired driving is not limited to alcohol or marijuana. There are many other substances that can affect driving and create risks on the road. I hope this project can help us develop a better understanding of marijuana-related impaired driving and how the cues we identify may compare with those associated with other substances.

If different substances affect driving and driver behavior in different ways, I think there is a lot of value in continuing this type of research in the future. I would be excited to have the opportunity to work with the team on similar studies involving other substances and continue building knowledge that can help law enforcement better identify impaired driving and ultimately improve road safety.