AUTOMATING BLOOD-PRESSURE MEASUREMENTS - THE DIVISION-OF-LABOR AND THE TRANSFORMATION OF METHOD

To what extent can computers undertake 'human' tasks, and what role do people play when computers begin to do 'their' jobs? This paper offer s an empirical analysis of the scope of medical computers in everyday applications, using, as a vehicle for investigation, a machine that is designed to detect and measure human blood pressure (BP). The paper d escribes experiments using 'novice' volunteers who, following written instructions, attempted to measure BP with a traditional sphygmomanome ter. Producing reasonable results proved extremely difficult: the proc edure is inherently complicated. It is clear that more experienced sta ff also deviate from the ideal instruction-based model: variations in techniques and results are widespread. But fieldwork shows that the ma chine, introduced to standardize the process, is successful: it produc es acceptable readings that are comparable over time. The paper invest igates how the machine is able to do this. I argue that the division o f tasks between human and sphygmomanometer, and between human and comp uter, has changed. The computer now performs what was previously a hum an task - but only a slice of the overall task. Furthermore, the compu ter uses a technique that is different from that traditionally used by people. The task is transformed during its automation. I suggest why some other tasks may be suitable for computerization, and why some tha t are not should remain essentially human, social activities.