In the world of data journalism, the advent of AI tools like ChatGPT has opened up new avenues for data analysis and visualization. Chad Skelton, a former data journalist at the Vancouver Sun and now an educator in journalism and data visualization, shares his insights into the capabilities of ChatGPT, particularly when enhanced by the Notable plugin for Jupyter notebooks.

Skelton’s exploration centers around a dataset detailing five years of bike thefts in Vancouver, encompassing nearly 7,800 incidents with detailed information on the date, time, location, and more. His objective was to showcase how ChatGPT, with minimal guidance, can generate meaningful analysis and visualizations from this dataset.

Initial Impressions and Setup

The process began with an overview of the dataset’s fields, such as date, time, and location, with ChatGPT providing metadata insights not explicitly present in the dataset. This initial step, including checks for missing values, laid the groundwork for deeper analysis.

Diving into Data Visualization

Skelton emphasized the importance of visualization in understanding data trends. ChatGPT generated various charts, including the distribution of thefts across different districts and times of the day. These visualizations revealed patterns such as higher theft rates in certain districts and peak theft times during evening hours. Interestingly, ChatGPT flagged a potential data reporting anomaly at midnight, suggesting a deeper investigation into reporting biases or data entry errors.

Further Analysis and Insights

Prompted for more detailed analysis, ChatGPT examined trends over time and the correlation between theft occurrences and various factors. This included a heat map showcasing theft occurrences by hour across different districts, offering nuanced insights into when and where bike thefts were most prevalent.

Seasonal Patterns and Anomalies

Upon querying about seasonal patterns, ChatGPT produced a chart highlighting a significant increase in bike thefts during the warmer months of July and August. This pattern aligns with intuitive expectations about biking frequency and theft opportunities.

Learning from the Data

One of the most striking revelations came from examining thefts reported right after midnight. ChatGPT suggested that a spike in reported thefts at this time might not indicate an actual increase in thefts but could be attributed to data reporting practices or entry errors. This insight underscores the importance of critical thinking in data analysis and the value of questioning anomalies.

The Educational Value of ChatGPT and Notable

Skelton points out that the entire analysis process, facilitated by ChatGPT and the Notable plugin, required minimal direct instruction. This aspect highlights the tool’s potential as an educational resource, allowing users to engage with data analysis and visualization through intuitive, natural language prompts.

Conclusion: The Future of Data Journalism with AI Tools

Chad Skelton’s exploration of bike theft data using ChatGPT and Notable showcases the evolving landscape of data journalism. AI tools are becoming increasingly capable of handling complex data analysis and visualization tasks, making them invaluable assets for journalists, educators, and data enthusiasts. As these tools continue to develop, their potential to democratize data analysis and enhance storytelling is immense, promising a future where in-depth data insights are more accessible than ever.

Source: here