Reducing the risk of aircraft collisions remains a primary airport safety issue. Evolution in air transportation brings new challenges including increased traffic levels, new information systems, and modified procedures, together with the requirement to maintain or improve safety. These challenges make it important to search for better ways to use automation to improve safety.Automation must go beyond single-level alerting to derive the greatest safety benefit from investments in surveillance and automation. At the same time, the automation must limit its demands made on flight crew attention, as unwanted outputs such as nuisance alerts can degrade performance and render such a system operationally unacceptable. A new concept for governing the behavior of an automated Conflict Detection and Resolution (CD&R) system called graduated intervention uses a control law that optimizes the time series of outputs ("interventions") given to an operator in terms of costs and benefits. A non-alert indication might be provided first, while an alert is provided only later if the situation continues to worsen. The approach is more flexible than a single-level alerting system because less disruptive interventions, such as nonalert indications, may be provided earlier while uncertainty and nuisance alert constraints prevent the use of alerts. This approach requires the CD&R system to anticipate potential conflicts as early as possible, based on expected aircraft behavior. A technique for representing predicted aircraft activities in a CD&R system for the purpose of generating early indications will be described.Prior work in the field of cognitive engineering supports the argument that operator response times fluctuate depending on a variety of factors, and that care in the design of the alerting system is needed to maximize the chances of a successful response. That design must address the need to present critical information to the operator in time to speed the process of assessing the situation and successfully resolving it. Providing a progression of information beginning with indications to aid situation awareness and potentially culminating with resolution guidance can provide an additional safety barrier against delay or error in responding to the situation. This paper analyzes the importance of response time and accuracy using a probabilistic model of both the ensemble of potential conflict situations and the spectrum of pilot responses. The effectiveness of graduated intervention is contrasted with single-level alerting using this model, focusing on quantifying the potential to reduce the rate of late or insufficient responses to a hazardous situation. The model is applied to a representative runway safety scenario. By predicting the time at which early information can be provided under the graduated intervention algorithms, and conservatively predicting the effect of that information on conflict response time, the model quantifies the benefit in terms of conflict outcome, measured across an ensemble of potential conflic...