In March of 1986, something phenomenal occurred at the East Texas Cancer Center. At the time, the significance of this event went unrealized, but as time passed, the implications became clear. Looking back, this event forever changed the way systems are designed and programmed.
The day began just like any other as a cancer patient was receiving his normal radiotherapy treatment from the Therac-25 machine. The machine had two distinct modes, “e” for electron and “x” for X-ray. Despite the obvious problems which could arise from having both an electron beam and a high powered X-ray beam programmed into the same machine, the operator assumed she could use the machine without encountering any problems. She began to input the prescription data, and due to her previous experience, she was able to input this information very quickly.
After double checking her inputs, the operator (a trained radiology technician) noticed that the mode was in X-ray instead of electron. Realizing the problem that this created, she hit the “cursor up” command in an attempt to correct the mistake. Then, after pressing the “return” key several times, the treatment was administered to the patient. Yet the technician was completely unaware that her actions had triggered a bug in the software.
During the course of the treatment, an error classified as “Malfunction 54” occurred, informing the technician that an underdose had been administered.
Following standard operating procedures, the technician hit “P” to proceed with the treatment. However, upon doing so, a second dose of treatment was administered to the patient, who immediately realized that something was wrong and began screaming. The operator, who was isolated in another room separate from the patient, heard nothing. There were audio and video monitors in the room which allowed the patient to interact with the operator, but that day, none of the equipment was working.
Five months later, the patient was dead. His death was directly attributed to the radioactive overdoses he received during his “treatment.”
So, this story is both sad and interesting, but what does it have to do with usability?
Usability deals with the ease of use (intuitiveness) of designs. More specifically, good design encompasses numerous factors, many of which were overlooked or ignored by the programmers of the Therac-25 system. Below is a list of what constitutes “good” design.
- Superior designs allow for the exploitation of human behavior. As cognitive beings, humans are natural pattern matchers and button pushers. If prior research or usability testing had been performed on this machine, test results would’ve revealed the rate at which an experienced user could input parameters. Yet after the recall of the Therac-25 machine, its makers discovered that the notorious “Malfunction 54” error would occur when keys were pressed successively under an 8 second time frame. An experienced user, I would imagine, would have the skill necessary to input such parameters at a rate that would fall below this time frame.
- Efficient designs ensure the clarity of the system through immediate visual clues. In the case at hand, there was no connection between the patient’s room and the operator’s room. Although audio and monitoring equipment existed, there was no backup system allowing for redundancy.
- Operative design provides informative descriptions of errors which occur in the system, allowing the user to gain not only an understanding of the malfunction, but a possible solution to correct it. “Malfunction 54″ is neither descriptive nor correct. Rather, the system should have specified that an underdose had been given, allowing the technician to identify the problem and improvise a solution. Had this occurred, the patient’s life could have been saved.
- Effective design creates a clear differentiation between two completely distinct modes of operation. An “e” and “x” design doesn’t provide any substantial visual feedback to the operator, as the letters are not physically displaceable. A better alternative would have been to create two different looking beam heads for the machine, allowing the operator to discern when they were using the wrong beam on a patient.
Take Aways
Mission critical system errors like those of the Therac-25 system illustrate the consequences of both bad design and the lack of user testing. In this case, the cost of poor usability was quite unfortunate; it literally became a matter of life and death. In fact, two other patients died from the same machine. Thus, in the end, we are left with some pressing questions: What is the cost of poor usability? How much is it really worth to you and your business?
