Speech Technology Magazine


Frequency of Use and Design

By Walter Rolandi - Posted Mar 8, 2004
Page1 of 1
Bookmark and Share

How Often and How Long
Before you can design an effective voice user interface, you had better answer the following questions:
• How often is the user likely to call the application?
• How long will the average interaction last?

Both questions are important, particularly when it comes to the ability of a persona to irritate a user but frequency (how often) is the more important of the two. When I teach my Master Class on VUI Design, I repeatedly stress its crucial importance in the design of an effective voice user interface and remind the attendees of Aesop’s admonition, “Familiarity breeds contempt.”

Imagine an insurance information system that someone might call once or twice a year. The system provides a great deal of information to the user and also allows him to perform a number of different tasks, often entailing four or more speaking turns. Given its complexity and varied abilities, coupled with the fact that most users are perpetual novices due to the infrequency of their calls, longer, more instructive prompts may well be in order. Because most users seldom use the system, design for the novice user is important in order to instruct the user and avoid confusion.

Now imagine a pizza takeout application that one invariably calls at least twice a week. From the user’s perspective, the system actually does only one thing: it collects some basic information. It greets the caller, determines the size of the pizza, the toppings, pickup or delivery and the method of payment. It’s a straightforward, form-filling application that can complete its task in as little as 20 seconds. An application such as this calls for short, fast paced prompts.

The Effects of Frequency
By the fourth or fifth time the user calls the insurance system (over a period of several years), some aspects of the system may seem familiar. However there is nothing particularly annoying about the length of the prompts or the explanations that the system continues to provide. They still seem helpful and instructive and may well contribute to a better user experience.

But by the fourth or fifth time the user calls the pizza takeout system (over a period of 15 days), everything that the system says and does has become utterly predictable to the user. Due to its relative simplicity and what I like to call its single-mindedness, the user quickly and easily learns how to use the application. The user knows exactly what the system is going to say and when it is going to say it. He learns to barge-in on the familiar prompts in order to speed things along.

The Intersection of Two Factors
These examples provide illustrations of the way frequency of use can affect design decisions and the resultant user experience. But the examples also illustrate a fundamental property of learning.

Basically any experience that is repeated becomes, in some way or another, learned. Again, the fellow wading his way through the insurance application once a year gains a sense of familiarity but that sense in no way resembles the predictive mastery that the twice-a-week pizza takeout user gains.

As a general rule, a system’s learnability increases as a function of the frequency of its use. Even a poorly designed system is easily mastered by its users if they have the misfortune of having to use it frequently. But in the hands of a good designer, frequency and learnability can ideally intersect for optimal results.

The question becomes: Which user do you design for — the novice or the frequent user? I am increasingly called upon to perform expert reviews that are basically exhaustive analyses of a VUI. I evaluate the VUI on the basis of three fundamental variables: feature functionality and value proposition; human factors and ease of use; and overall production quality.

A common finding is that system designs fail to adequately address the novice or the frequent user question. Among frequently used systems, prompts are usually too long and far too much information is conveyed to the user as to what the system is, what it does and how the user should use it. Among infrequently used systems, there is often insufficient state testing logic behind the application to intelligently discern when a user is having problems.

By now, most of us have made use of an automated checkout machine at a grocery store. Interaction modality is akin to DTMF in that the machine speaks to users but they must respond by touching screens or entering numbers into a keypad.

It is interesting to watch people use these machines and observe the opposite ends of the user spectrum. New or infrequent users often become confused. They are unfamiliar with the task at hand and usually get out of sync with the machine. Frequent users, on the other hand, know exactly what to do. They completely ignore and often cut short the cheery instructions that the machine invariably emits. They are one step ahead of the machine, all the way out the door.

These checkout applications would be better if they didn’t speak at all, but they provide an example of a one-size-fits-all design that fails to address user frequency and ends up annoying almost everyone. What’s worse, designs like this train their most experienced users to ignore what they say and, ironically, demonstrate a superfluous use of speech.

Dr. Walter Rolandi is the founder and owner of The Voice User Interface Company in Columbia, S.C. 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.

Page1 of 1