Characterization or characterisation is the representation of characters (persons, creatures, or other beings) in narrative and dramatic works. … This representation may include direct methods like the attribution of qualities in description or commentary, and indirect (or “dramatic”) methods inviting readers to infer qualities from characters’ actions, dialogue, or appearance.
— “Characterization.” Wikipedia, 26 Jan. 2025. Wikipedia, https://en.wikipedia.org/wiki/Characterization.
Data analysis in a vacuum is pointless without context. Context typically consists of several components: driver, vehicle, track, and environment. Environment is a much larger, more complex ball of twine as it could encompass weather, race format, and other competitors; for the time being let’s not dive into that aspect of our race context.
Now that we have a context defined, how do we know if something is similar to what has been seen before, or if it’s noticeably different? Is the driver bringing their A-game this weekend, or a bit off for some reason? Is the vehicle running nominally, or is something, “just not right?” Is the track absolutely perfect (dry, warm enough but not too hot, clean, smooth, etc) or is something different from other tracks – or even the same track on a different day?
While data analysis with context is important, context without baselines (or ground truths) is equally useless. What is the baseline for a driver’s performance look like? How about that Spec Miata MX-5; how is its baseline different from a Ligier P3 car? And how do you compare a given track day-to-day, much less against another track? This is where characterization comes into play.
Characterizing a driver might take into account driving style, temperament, current standings, or many other aspects of what might impact their performance. Similarly a car could be characterized mechanically, but also how it performs under a range of conditions or responds to changes over time; it is after all a physical structure that evolves and wears down through repeated usage. We hear about track characterizations all the time from experienced drivers and coaches discussing the, “personality,” of given circuit, but what if we could quantify that as well?
By developing characterization approaches for drivers, vehicles, and tracks, we can better ascertain where performance is being lost in a given session by identifying if each major contributing component is playing its part to perfection, or is truly out of character and needs some attention. Identifying these baselines and recurring patterns is what AI does best; this is only part of what we’re training RAiCE agents to do.
