Limitations of Technology | Voice Recognition Software - I Robot. Not!
In I, Robot, Isaac Asimov’s famous collection of science fiction stories, he explored various facets of robotic technology and how it might affect humanity. In Asimov’s stories, robots were governed by laws designed to protect humans, but events in the stories often went awry because of strict, unwavering robot obedience, devoid of human understanding and judgment.
I am reminded of these stories every time I field a phone call from a student who is upset because someone said that medical transcriptionists are being replaced by voice recognition software—because it’s the same idea.
Developers and salespeople of Voice Recognition (VR) technology are, understandably, proud of their invention and enthusiastic—perhaps overenthusiastic—about its abilities and applications. Office managers whose main concerns are to keep track of personnel and cash flow of a medical office, and even doctors who don’t really understand what happens to their dictation between the time they speak it and the time they see it in the patient record, are excited with the prospect of a machine to take over the task of medical documentation without expecting a paycheck. So, the salespeople pitch to doctors’ offices and a Voice Recognition (VR) system is purchased (against a very large price tag) based on the idea that a machine is superior to a human, that a computer program will deliver perfect compliance and accuracy sans the hassles and expenses of dealing with a person—the same pitfalls explored in Asimov’s stories.
As trained professionals in the medical transcription field, it’s easy to see the error in these assumptions. Why is a computer unable to handle the entire task of medical transcription? On the surface the task appears simple: to accurately turn spoken words into a text format. This seems like it would be appropriate for voice recognition software, right? Well, no, because the task is really much more than just robotic typing in response to audio. Trained medical transcriptionists use uniquely human abilities to make decisions, read context, accept variation, and understand meaning. Medical transcriptionists leave out all the ‘ums’ and ‘uhs’ as well as comments the doctor makes to his kids or passing colleagues that have nothing to do with the report; medical transcriptionists follow verbal directions and dictated backtracks to correctly insert information spoken out of order; MTs also can work with charts and lists if necessary, inserting headings and subheadings as needed; MTs organize and rephrase at times. In short, trained MTs ensure that what shows up in the patient’s medical report reflects what is meant, not just what is said. This demonstrates critical thinking—a uniquely human ability. The cold, hard fact is machines lack the ability to think critically.
In a presentation given to the 8th Annual International Conference on Critical Thinking and Education Reform, Summer 1987, Michael Scriven & Richard Paul put forward these simple ideas:
1) A set of information and belief generating and processing skills, and
2) The habit, based on intellectual commitment, of using those skills to guide behavior.
It is thus to be contrasted with:
1) The mere acquisition and retention of information alone, because it involves a particular way in which information is sought and treated;
2) The mere possession of a set of skills, because it involves the continual use of them; and
3) The mere use of those skills (“as an exercise”) without acceptance of their results.
If you apply these defining statements to contrast and compare human medical transcriptionists and computer software, it’s easy to see the strengths of the human transcriptionist.
Medical dictation presents a great deal of variation in speech pattern and voice modulation, accents, regional dialects, etc., from doctor to doctor; although voice recognition software is advertised as having the ability to “learn” a particular person’s speech habits, this is limited and requires time and patience on the part of the person speaking to train the machine. Also, the “training” accomplished for one person’s voice doesn’t help the next person who uses it. The machine, therefore, doesn’t really have a habit, or an intellectual commitment that guides behavior, like a trained human transcriptionist does; all it has is the ability to acquire and retain information.
There is a very large task of making sure the information makes sense. A machine cannot take verbal cues, apply them to the constructs of the language, make logical allowances for variation, draw correct conclusions and accurately convey meaning. Again, there is no real processing of information, only the application of acquired skills as an exercise without (and this is the important part) acceptance of their results. A machine doesn’t know, nor does it care if it makes a mistake. In contrast, a person, especially a well-trained professional in the field, accomplishes all this naturally and instantaneously. And a person cares.
Ironically, it is also human nature to defend a decision, especially if a significant investment has been made. Medical offices that convert to a “complete” voice recognition system are very motivated to implement it and the office personnel work very hard to turn the excitement into reality. Some medical offices seem to be an acceptable fit; they have fairly modest demands and repetitive documentation needs, so after a period of adjustment involving changes in the doctors’ habits and a likely dip in productivity, the VR system settles in as an acceptable, although not superior, option. Other office and hospital settings are not conducive to the limitations of a VR system at all; the reality in this case is the office must now work with a system that requires a great deal of adjustment, may very well put more demands on the existing office staff as well as the doctors, and does not really work any better than their old system—and they’re still making payments. Some offices grit their teeth and make it work; others realize the system is good for a first draft and hire MT editors; some offices simply give up and call back their transcription staff.
Technology will continue to develop, of course. Perhaps someday our lives will resemble the pages of science fiction stories in our easy use of computer intelligence designed to enhance our daily existence. Keep in mind, however, that as thinking, feeling, analyzing human beings, we are irreplaceable. No technology approaches the unique abilities of a person. Just as Asimov’s stories illustrate, the human component is vital to the success of any endeavor involving the welfare of people—like medical transcription. Remember this, and value yourself. We are not robots; we are so much more!