With the advent of connected and automated vehicles, naturalistic traffic modeling is becoming increasingly important for the purposes of congestion control via cooperative driving and intelligent traffic management. The focus of this paper is to develop a data-driven traffic interaction model which can help in the development of multi-agent automated driving control schemes that mimic and blend with other human-driven vehicles. To that end, a probabilistic data analysis approach is used to derive an energy function that describes interactions between vehicles on highways. These interactions arise from the psychophysics of humans driving the vehicles. The analysis suggests the existence of a simple interaction law between human-driven vehicles that is based on the expected time it takes for the vehicles to collide. The approach taken in this paper helps in analyzing probable states that the individual vehicles in the traffic would have, thereby facilitating the development of intelligent traffic management tools that account for individual vehicle states.