1. Human Factors Progress IDS Project June, 2004
Nicholas Ward
Mick Rakauskas
Jason Laberge
Janet Craeser
HumanFIRST Program

2. Human Factors Tasks Analyze problem Simulate case site
Task analysis
“What are the task elements of crossing an intersection?”
“Where in this sequence in the task failing?”
“Who is most at risk?”
Information analysis
“What information supports task behavior?”
“Which information is misused or missing?”
Simulate case site
Propose interfaces and simulate candidate
Review previous solutions
“What has not worked before?”
Evaluate candidate interface

3. Task Analysis Detect intersection Decelerate and enter correct lane
Signal if intending to turn
Detect and interpret traffic control device
Detect traffic and pedestrians
Detect, perceive, and monitor gaps
Accept gap and complete maneuver
Continue to monitor intersection

4. Target Population
Older drivers (> 65 years) have a high crash risk at intersections
Drivers > 75 years had greatest accident involvement ratio (Stamatiadis et al., 1991)
Drivers > 65 years 3 to 7 times more likely to be in a fatal intersection crash (Preusser et al., 1998)
Drivers > 65 years over-represented in crashes at many rural intersections in Minnesota (Preston & Storm, 2003)

5. Abstraction Hierarchy
Decision Support Systems (DSS) for drivers can convey different types of information in a number of forms. Determining what information to present as well as when and how to present it can be a challenging design problem. Ecological Interface Design (EID) can help by identifying not only the information content but also the optimal form (Lee et al., 2003). EID uses the abstraction hierarchy (AH) and the skills, rules, knowledge (SRK) framework to identify environment and operator constraints that are relevant for display design (Figure 1). Constraints govern the manner in which the human-machine (operator and the environment) system interrelate and can be either limiting conditions (i.e., weather) or overall system goals (i.e., safety). The AH identifies environment constraints and the information elements that can be used to convey the constraints to operators. In this respect, the AH helps identify the information content that could be presented in an ecological interface (Figure 9). The SRK taxonomy identifies operator constraints by focusing on basic categories of human performance. This analysis helps ensure the information content is represented in a form that is consistent with operator performance and information processing limitations (Lee et al., 2003).
The abstraction hierarchy (AH) is a framework that helps identify the environment constraints for an ecological interface (Lee et al., 2003; Rasmussen, 1983; Vicente, 2002). The AH determines the interface content because the environment constraints are made visible to the operator via one or more information elements in the display. In an AH, environment constraints are categorized into five levels of abstraction and three levels of detail. Means-end relationships connect each level of abstraction and whole-part relationships describe how constraints are connected at the detail level. The functional purpose level describes underlying goals of the system, such as operator (or driver) efficiency. The abstract function level lists constraints and principles that need to be satisfied to achieve the goals identified at the functional purpose level. The balance of performance (i.e., average speed) and cost (i.e., mileage) is an abstract constraint that affects the functional goal of operator efficiency. The general function level identifies standard processes and features that are the means to influence the constraints at the abstract level. Examples of constraints at this level include general traffic dynamics such as elasticity and other driver intent, both of which affect the performance and cost balance and overall efficiency. Physical elements of the system, including their relationship to each other and the environment, are provided at the physical function level of analysis. This could include the road type and weather conditions. Lastly, physical form is where the appearance, anatomy, form, location, etc. of specific elements identified in the physical function level are listed. A stalled car in the left lane of I-494 just before the I-35W interchange would be an example of a constraint at the physical form level.
EID is different from traditional task analysis because it focuses on the entire work domain, but emphasizes environment constraints. EID also supports operator tasks at all three levels of performance and provides information at more than one level of abstraction. In contrast, traditional task analysis focuses only on known tasks and therefore may not identify information elements that operators need when they encounter novel situations. Traditional analyses also often represent information at only one level of detail. Therefore, EID encourages skill and rule based performance but also supports more effortful knowledge based performance when needed (Vicente, 2002). This could result in a more detailed and accurate internal representation of the system and higher operator trust in the resulting ecological interface (Lee et al., 2003).

6. Generic Support Intersection / Control device Vehicle presence
Vehicle speed, distance, time
Size of gap in traffic
Safety margin of gap
(specified location in traffic)
Most prior systems have been limited to:
Emphasize presence of intersection and traffic control device.
Presence of approaching cars.
Approach speed of cars.
Given that awareness of intersection and compliance with traffic control devices is not the problem in our case, this method (1) will not benefit safety.
To the extent that drivers are at risk because of problems with more complex information needs (C and D), simply presenting information about vehicle detection will not benefit safety.
Because the research does not give evidence of the relative importance of these factors toward crash risk, it is necessary to design options for ALL of these needs. Note also, that the highest level (D) also satisfies the lowest level (A), but not conversely.

7. Minnesota Context
In Minnesota, most drivers stop before proceeding (Preston & Storm, 2003)
57% stopped in 2296 rural thru-STOP accidents
87% of right angle crashes at US 52 and CSAH 9 occurred after the driver stopped
NOT a violation problem
Instead, a gap acceptance problem
Detecting vehicles (speed, distance, time)
Perceiving gap size (and location)
Judging safe gaps
Implies that sign violations are not the primary crash factor for rural thru-STOP intersections in MN
Sign violations occur because of problems with divided attention, distraction, inattention, and motivation
Gap acceptance crashes occur primarily because of problems with gap detection, perception, or decision-making

8. Minnesota Location

9. Road Network

10. Intersection

11. Elevation

12. Crossing

13. Demonstration

14. Interface Task Design Tenets Expert panel review of concepts
Prohibitive (not permissive).
Decision remains with driver.
Design for worst case.
Use MUTCD sign guidelines.
Consider diverse range of option rather than refine a concept.
Expert panel review of concepts
Everyone had own perspective.
No consensus for best sign.
Some signs ejected.
Interface demonstration
IDS TAP
MN Pooled fund
MUTCD
Revised design
The problem of intersection crashes involved a state of the art review of crash data and literature.
A task analysis was completed of the tasks involved in safely getting through an intersection. This identified what actions needed to be taken.
A information processing model of the driver was devised to specify what information is needed to support these actions.
An Abstraction Hierarchy is a formal method of organizing information elements in a task environment to support the design of an ecological interface.
Together, this process generated interface concepts to provide one or more types of information to the “stopped” minor road traffic using only prohibitive formats:
Detection of approaching mainline vehicles.
Presentation of approach speed and arrival time of vehicles.
Presentation of gap size
Specification of a safe gap
We need to present at more than one of these levels of information because the available data does not definitively indicate exactly where in this information process crashes are occurring.
For example, if we only adopt a simple and cheap solution that detect vehicles, then it may not reduce crashes is the real problem is with estimating safe gaps.
Note that detection of the intersection and compliance with the stop sign is NOT a problem for our case study.
Note also that these interface designs were generated by the human factors analysis are not by MUTCD guidelines (although the coloring and placement of signs are consistent with these guidelines).
This was done to generate a new tool set of candidate interfaces that extend beyond current MUTCD options.

15. Four Prototypes Hazard Beacon Flashing sign activates when
intersection is unsafe.
System tracks
arrival time
(or speed)
of lead vehicle
Speedometer
Speed monitor
for lead vehicle.
Flashes red when
near or far-side
vehicle is speeding.
Hybrid
Arrival time
countdown for
lead vehicle.
Prohibitive
symbol relative to
maneuvers based on
near and far-side
traffic conditions.
Spit-Hybrid
Median position
with logic for North
Left nearside
position for North
and South.

16. Baseline

17. Hazard Beacon

18. Speedometer

19. Hybrid

20. Split Hybrid

21. Conclusion Task Completed: Intersection selected
and simulated with high
Geospecific accuracy.
Task Completed:
Interface concepts generated.
Task on schedule:
Experiment outlined.
Interface logic tested.
Traffic models under review (gaps).