Repeat or Not Repeat

The Common Assumption
What is the proper role that repetition should play in a voice user interface? This question frequently arises when designing a VUI, particularly if the VUI is intended to simulate "natural speech" or "conversational dialog". The common assumption is that repetition is bad because it doesn't sound natural and it occurs only infrequently in human-to-human conversation. An Uncommonly Known Fact
It is simply untrue that repetition occurs infrequently in human-to-human conversation. The fact is, we all constantly repeat what we hear others say. Famed psychologist B. F. Skinner noted this phenomenon and coined the term "echoic" to signify a particular incidence of its occurrence. Basically, an echoic occurs whenever a speaker repeats ("echoes back") all or part of something someone else has said. Consider the following example: "I'd like a cheeseburger and a beer".
"Ok. What would you like on your cheeseburger"?
"Just mustard".
"Ok. A cheeseburger with mustard only. Would you like some fries"?
"Sure, why not?"
"And what kind of beer"?
"What are the choices"?
"Draft or bottled…"
"Draft". Each italicized word in the example is an echoic and the role that they play in a conversational exchange should be apparent. Among other things, they serve to confirm what the listener heard the speaker to say and to establish the pace for conversational turn taking. Would anyone deny that the example constitutes a conversational dialog? Is there anything particularly unnatural about it? Stuck in IVR Hell?
Many designers, in seeking to create dialogs that seem conversational or natural, attempt to eliminate or, (more accurately speaking), to disguise repetition. The goal is often expressed in terms of wanting to prevent the user from feeling "stuck in IVR Hell". This practice can take several forms. For example: Variable prompts How can I help you? (first prompt played)
And how may I help you now? (second prompt played) Unique or escalating recognition error messages I didn't quite get that.
(error 1)
I didn't quite get that either.
(error 2)
I'm sorry but I'm afraid I didn't get that one either. (error 3) While these practices can superficially appear to be more natural or conversational, particularly to a first time user, one wonders how designers can convince themselves that they will make a difference to the user experience over time. No one wants a user to feel that they have fallen into IVR Hell or that they are "stuck in a loop". Yet in most cases where variable prompts are employed, the user is, in fact, just that: stuck in a loop. The system may support Cancel, Exit, Start Over, Main Menu and many other ways to escape. But the fact remains, in almost all cases where variable prompts are currently employed, the user's choices for forward movement are severely limited. What this ultimately means is that the user, especially over many trials, will learn to ignore the individual differences among the variable prompts in order to concentrate on saying precisely what is necessary to move himself forward. Ironically, less variable prompts can often serve to get the user more quickly back on track. The Coveted Goal and the Reality of Practice
Conversational dialog may well be the goal, but current practice seems to suggest that until our goal is reached, we should in the meanwhile pretend that we are already there. Users are expected to play along in what amounts to a fanciful game, a game in which they are obliged to make believe their conversational partner is capable of continuously emitting novel utterances. No existing application can truthfully claim to appear continuously novel. (If any reader can prove otherwise, please contact me by email, wrolandi@wrolandi.com). What this means, in practice, is that the simulated novelty of variable prompting schemes will eventually (and probably rapidly) be lost to the repeated user. Repetition is the friend of the wise designer. Not only is it the basis of all learning, using it serves to limit the expectations of the user by reminding him that he is talking to a (stupid) machine. Repetition implies predictability and predictability assures the user that he is in control. These are not bad messages to send to your user. Or do I need to repeat myself?

Dr. Walter Rolandi is the founder and owner of The Voice User Interface Company in Columbia, SC. Dr. Rolandi provides consultative services in the design, development and evaluation of telephony based voice user interfaces (VUI) and evaluates ASR, TTS and conversational dialog technologies. He can be reached at wrolandi@wrolandi.com

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues