ISA Trial, US
|ISA Trial, US|
|Type: Field operational test|
|Tested system/service: Intelligent Speed Adaptation|
|Countries: USA||50 test users|
|? partners||8 vehicles|
|Active from 2011 to 2011|
|Ian J. Reagan|
|National Highway Traffic Safety Administration|
|Data catalogue||Tools catalogue|
|Data sets used in this FOT:
No data set is
|The following tools|
were used in this FOT:
No tool is linked
Speeding is a factor in a significant number of traffic fatalities. The systems tested in this project have been evaluated outside but not within the United States. These studies indicated that similar systems led to reductions in speeding. In this field experiment, the authors tested an alerting system and a monetary incentive system with the objective of reducing speeding more than 5 mph faster than the posted speed limit. For this study, eight vehicles were instrumented such that vehicle speed and speed limits were linked in real time. A total of 50 participants drove assigned vehicles for 4 weeks. Week 1 was a baseline period; during Week 2 or Week 3, 40 participants experienced the alerting system that issued auditory and visual advisory signals when drivers exceeded the limit by 5 mph or more. Of these 40 individuals, 20 experienced the monetary incentive system during Weeks 2 and 3; Week 4 was a return-to-baseline period. A control group of 10 drivers experienced neither system during the study.
In a proof-of-concept study, Brookhuis and de Waard (1999) demonstrated the potential benefit of an early advisory ISA system by having drivers complete baseline and ISA drives on a fixed 35-min route with five speed limits. The system provided graded visual and vocal feedback to drivers when they exceeded the limit. The gradations of visual feedback were green to denote adherence to the speed, yellow to indicate intermediate violations, and red to signal violations greater than 10 km/h. The vocal feedback coincided with the red visual display. The researchers showed a significant reduction in the time that drivers exceeded the limit by 10% when driving with the ISA system. An obvious limitation of this study was the contrived nature of the drives: Participants drove the same route and there was no basis to assess the effect of the system over time.
However, with advancements in GPS technology have come long-term field operational tests (FOTs) of ISA systems in more naturalistic settings. To date, the most comprehensive ISA FOT occurred in Sweden between 1999 and 2002 (Biding & Lind, 2002). Biding and Lind (2002) report the results of their large-scale test of four systems in four urban areas across a 2-year period. Two systems tested were exclusively advisory-level ISA. In two of the four sites, experimenters recorded baseline driving measures during the 1st month of the trial then activated the ISA systems and continued data recording for approximately 2 years. The analyses compared the pre-ISA activation period with two ISA activation periods of 1 month each. Overall, Biding and Lind reported reductions in speeding that ranged from 10 to 20 percentage points across sites and speed limits when the ISA system was active. The authors noted that the reduction in speed violations was attenuated during the second post-ISA period although still significantly lower than the baseline period.
Original title of the report: "The Effects of External Motivation and Real-Time Automated Feedback on Speeding Behavior in a Naturalistic Setting"
Details of Field Operational Test
Start date and duration of FOT execution
Hypotheses. On the basis of previous ISA research (e.g., Biding & Lind, 2002; Harms et al., 2007; Jamson, Carsten, Chorlton, & Fowkes, 2006) and on principles of behavior theory (Eisenberger & Cameron, 1996), we predicted that the combination of MI and automated feedback (AF) would result in the greatest reductions in exceeding the speed limit compared with either MI or AF alone and with baseline and control conditions. We also predicted significant reductions in speeding as a function of individual effects of AF and MI compared with baseline and control conditions.
Experimental design. We assigned 50 drivers randomly to three MI groups. Of these drivers, 10 acted as a baseline group, driving for 4 weeks with no AF or MI. Another 20 drivers received no MI but received AF during either Week 2 or Week 3 (counterbalanced). The remaining 20 drivers received MI during Weeks 2 and 3 and received AF during either Week 2 or Week 3 (counterbalanced) (see Table 1). Weeks 1 and 4 served as baseline and reversal periods for the 40 experimental participants, respectively.
This design builds on the work of Battista, Burns, and Taylor (2010), Harms et al. (2007), Hultkrantz and Lindberg (2003), and Mazureck and van Hattem (2006) by including a control group and a reversal (return-to-baseline) period. In contrast, Battista et al. and Mazureck and van Hattem used an ABA design with no control group, and Harms et al. and Hultkrantz and Linberg used an AB design with control group. In addition, the current project and the research completed by Harms et al. were the only designs to isolate the effects of the advisory speed information. Additionally, the design of the advisory system was a graded alert, and this design feature was based on participant feedback from Biding and Lind (2002), wherein participants indicated that an alert that distinguished between different levels of speeding violations was preferable to a binary alert.
Results indicated that the incentive system resulted in significant reductions in driving faster than the posted limit, and the feedback system led to modest changes in speeding. In the condition in which drivers experienced the feedback and incentive, reductions in speeding were similar to those found during the incentive-only condition.
Summary, type of funding and budget
National Highway Traffic Safety Administration
Cooperation partners and contact persons
- Public Authorities:
- Vehicle Manufacturer:
- Research Institutes:
- Others (specify):
Applications and equipment
Project staff instrumented eight vehicles for use during the field study. NHTSA provided a 2002 Oldsmobile Intrigue, a 2001 Saturn L 200, a 1998 Chevrolet Malibu, a 2000 Ford Taurus, a 2005 Cadillac STS, a 1999 Toyota Camry, a 2003 Toyota Corolla, and a 2004 Toyota Sienna.
Equipment carried by test users
MI system. The MI condition was structured as a bonus system with a delayed incentive and an immediate disincentive. Individuals in the MI condition began Weeks 2 and 3 with $25.00. In a manner similar to Harms et al. (2007), the bonus declined by 3 cents every 6-s period that the driver remained 5 to 8 mph faster than the limit. The penalty increased to 6 cents if the driver was 9 mph or more faster than the limit during any segment of the 6-s period. A visual display, analogous to a meter in a taxicab, provided updated bonus amounts but displayed the information only when the ignition was turned on or off.
Advisory display. The display box that presented the updates about the incentive also displayed the visual speed alert and housed the speaker that annunciated the auditory component of the alert. The auditory alert included two 400-Hz tone stimuli to advise drivers during the AF conditions. The research assistant ensured that the alerts were audible from the driver’s seat in the presence of the ambient noise of popular music playing at a level deemed to be “loud.”
The temporal pattern of the auditory alert varied as a function of the magnitude of speed violation (Alert A for 5 to 8 mph past the limit and Alert B for 9 or more mph past the limit). The graded alert builds on and stems from previous ISA FOT results (Biding & Lind, 2002). Specifically, Biding and Lind’s (2002) participants stated that the binary alerts they experienced could be improved by presenting alerts that differentiated between levels of speeding. Each alert in the current study lasted for 3 s. Alert A consisted of four bursts with two pulses per burst; Alert B consisted of four bursts with three pulses per burst.
Speed map. Researchers obtained blueprints of the study areas and associated speed limits from the local governments. The area mapped included portions of Kalamazoo County, which is 573 square miles and has 1,263 miles of roads as well as the major arterials that flow into and out of Kalamazoo and Portage cities. Approximately 730 miles of roads were mapped for this study, and approximately 80% of the driving that participants completed during the experiment was on the mapped roads. With regard to mapping speed transition points, researchers noted the distance between a transition point and the nearest intersection and then transposed this information to the color-coded map so that the transition point in the database was accurate within 50 feet of the speed limit sign. Persentech, Inc., integrated the speed limit information into an existing Automate™ GPS device. The project’s software engineer designed the microprocessor to receive GPS and speed limit information from the GPS device and vehicle speed information from the antilock brake sensors or vehicle speed sensors. The engineer filtered out instances of zero speed during the data reduction process. From this input, the microprocessor recorded driving data and activated the incentive and feedback systems.
Pre-simulation / Piloting of the FOT
Method for the baseline
Techniques for measurement and data collection
Pilot testing. Throughout the development of the instrumentation and prior to full-scale data collection, we pilot tested components of the ISA system and microprocessor to ensure the functional reliability of each, to measure and improve the validity of the speed limit database, and to assess the combined effects of MI and AF on 3 pilot participants who were naive to the study. These iterative procedures included working with the GPS manufacturer to change speed limit and latitude and longitude values when we noted inaccuracies during field test drives. We adjusted the AF and MI parameters on the basis of daily and weekly traffic flow conditions on the primary, secondary, and arterial roads in the test area as well as previous research (GHSA, 2005; Harsha & Hedlund, 2007).
While on these drives, the research team noted the speeds of the prevailing traffic and whether and to what extent drivers had the opportunity to speed. Additionally, the research assistant drove for 1-week increments assuming the behavior of a driver who would drive the speed of a platoon of vehicles and otherwise moderately exceed the speed limit when the opportunity existed. We computed dependent variables (DVs) from the raw data files and tested reliability with the summary data files (all r values > .99). The effects of the AF and MI on the pilot test participants were encouraging.
Week 1. Participants received the vehicle at the beginning of Week 1. At that time, the experimenter informed drivers that the study was testing an emerging traffic safety system and that the vehicles had systems that recorded distance traveled, speed, seatbelt use, GPS, and time of day.
Participants provided a second informed consent and answered self-report questions. The experimenter provided participants an overview of their assigned vehicle and instructed participants that during the trial, they should drive as they would during normal, everyday driving. Participants were aware that a number of safety-related driving behaviors were recorded but were not specifically told that the target behavior in the study was speeding.
Weeks 2 and 3. After Week 1, participants who met distance and speeding exposure criteria continued to Week 2. The distance criterion was to drive approximately 100 miles during Week 1. To ensure that participants represented habitual speeders, the speeding criterion was based on the bonus structure used for MI participants. Therefore, only participants who would have lost approximately 35% of the bonus amount ($8.00) during the baseline week were allowed to continue. We established these criteria because the primary focus of the study was to assess whether the treatments affected speeding. The experimenter met with participants who satisfied the criteria to provide further instructions and had participants fill out self-report questions about sensation seeking and automation use and complete the NASA–Task Load Index (NASA-TLX) for Week 1. To aid in their completion of the NASA-TLX, the researcher asked participants to consider mental workload demand associated with the overall driving task. Instructions for Weeks 2 and 3 included explanations of the AF and MI systems, depending on group assignment. The 40 participants in the MI and no-MI groups drove for 1 week, either Week 2 or Week 3, with the AF system in active mode. The researcher used stratified random assignment for this condition to ensure that 10 participants in each MI condition experienced active AF during Week 2 and that the remaining 10 per group drove with AF active during Week 3.
For participants experiencing MI or AF, the researcher used a street map to show where the roads were mapped and explained that the system would “hibernate” if they ventured outside the area. Participants completed ratings of trust and acceptance of the MI and AF systems at the end of any week they drove with the systems. At the end of Weeks 2 and 3, the participants in the MI condition provided trust and acceptance ratings on the MI system. The full sample of 50 participants completed the NASA-TLX at the end of Weeks 2 and 3.
Week 4. During the final measurement period, Week 4, participants drove their assigned vehicles with AF and MI systems deactivated, as in Week 1. At the end of Week 4, participants provided subjective workload ratings via the NASA-TLX and then completed a debriefing interview.
Recruitment goals and methods
Participants were a convenience sample of 50 licensed drivers (26 males and 24 females) with at least 5 years of driving experience who lived and worked in the Kalamazoo, Michigan, area. Table 2 presents the number of males and females and average age for the sample. Drivers convicted of impaired or reckless driving or who had their license suspended were prevented from participating.
Researchers followed a multistage recruitment process. Potential participants signed an initial informed-consent document and provided self-report information about driving habits and driver license numbers. Drivers who met safety and driving exposure criteria signed a second informed-consent document. Drivers received compensation for their participation in the field study. The test vehicles received a full tank of gas on Day 1 of the experimental trial, and participants received $80 for completing experimental activities.
Methods for the liaison with the drivers during the FOT execution
Methods for data analysis, evaluation, synthesis and conclusions
Sources of information
Final report can be found at: http://hfs.sagepub.com/content/55/1/218.full