ON THE RELATIONSHIP BETWEEN HCI AND TECHNOLOGY ACCEPTANCE RESEARCH - Human-Computer Interaction and Management Information Systems: Foundations - page 379

of behaviors, and should be applicable to computer usage behavior. Second, there was a growing
awareness in the business and management literature about the importance of using valid and reliable
measures of constructs in order to successfully operationalize theoretical models (Churchill, 1979).
Could the lack of precise theoretical definitions of attitudinal constructs, coupled with the unknown
reliability and validity of measurement instruments, be to blame for the mixed and inconclusive track
record of MIS attitudes as of the early 1980s?
Whether looked at from the perspective of an organization deciding to build or purchase a new
software application or from the perspective of a vendor organization developing a commercial
software product to meet customer needs, a key set of decision variables under the control of prac-
titioners is the functionality and interface design characteristics of the target system. I reasoned that
functionality characteristics should be chosen to maximize the perceived usefulness of a system to
users, and interface characteristics should be chosen to maximize a system's perceived ease of use
among users. TAM defined perceived usefulness as a potential user's expectation that using the
system would improve his or her job performance, and perceived ease of use as the perception that
using the system would be free of effort. TAM hypothesized that perceived usefulness and ease of
use would play the role of causally linking design choices to users' intentions to use a new system.
TAM further hypothesized that because usefulness was a superordinate goal for most people, it
would therefore be more strongly linked to usage intentions than ease of use.
USER ACCEPTANCE TESTING: BEYOND USABILITY ENGINEERING
TAM was formulated not only to provide theoretical understanding of how system design charac-
teristics influence user motivation through their effects on perceived usefulness, perceived ease of
use, and intention to use, but also to provide the basis for a practical and effective “user acceptance
testing” methodology for predicting the degree of user acceptance of a new system based on meas-
ures from users who had limited exposure to an early prototype of the system. For such user accept-
ance testing to be practical, I needed to establish valid and reliable measurement scales that were
convenient to use in practice and were predictive of user adoption behavior in the workplace.
One particular study stimulated my thinking about the promise of early prototype testing. Gould,
Conti, and Hovanyecz (1983), working at IBM, sought to understand the usability of “listening
typewriters” based on speech recognition. Speech recognition technology was not yet sufficiently
mature to provide a working prototype for user testing, but extensive research and development was
under way, and speech recognition was expected to become commercially viable within five years.
IBM needed to understand the usability impact of specific system design features such as vocabu-
lary size, error rate, and continuous vs. isolated-word speech in order to guide research and devel-
opment investments for speech recognition technology. Gould et al. (1983) simulated speech
recognition by feeding users' dictation to expert typists hidden in the next room. Software was used
to control the vocabulary size and error rate of concurrently transcribed text, giving users the illu-
sion that the words appearing on the screen in front of them were being produced by speech recog-
nition technology. Gould et al. (1983) measured the typical HCI metrics of task completion times
and error rates, comparing the listening typewriter to alternative dictation technologies. Although
their study included a single-item measure of satisfaction, the measure had unknown psychometric
properties and unknown ability to predict actual user acceptance of speech recognition systems with
particular features. My research on TAM sought to develop measures that could be used in such sit-
uations to provide a valid indication of the relative degree of user acceptance of each of the design
alternatives. Just as Card, Moran, and Newell's (1980) keystroke-level model sought to allow
designers to predict the objective usability of a system's user interface from the design of the