Human Factors in Voice Interface Design Jeff Dworkin Segment Marketing Manager [email protected].

Human Factors in Voice Interface Design

Jeff Dworkin

Segment Marketing Manager

[email protected]

www.dialogic.com

Company Confidential • © Copyright 2008 Dialogic Corporation. All rights reserved.

Slide 2

Extends Mobile VAS Segment Leadership

Video Algorithmic and Analytics Leadership

Extends Technology Enabling MSS Leadership− Fax Segment MSS Leadership

Converged Communications Technology Enabling Market Segment Share Leadership− Dialogic “pioneer” history, relationships and patent portfolio− Enterprise Gateway

Established SS7 / Signaling Part of Business

Established HMP as core to Dialogic customer value proposition

Deeper Service Provider Segment Products / Customers− Service Provider gateway and IP media server

Extends Technology Enabling MSS Leadership

TDM to IP Transition Leadership Extend into Web communication and

open source ISV innovators Enabling Video IP Streaming Value Added

Services

2006 2007 2008

“VIDEO IS THE NEW VOICE”™

Dialogic Evolution

“VIDEO IS THE NEW VOICE”™

Mission: To Enable Secure Multimedia Communications Through Any Network To And From Any Endpoint In The World

www.dialogic.com


Slide 3

What is Human Factors?

Ergonomics – an applied science concerned with designing and arranging things people use, such that they interact most efficiently and safely.

Ergonomics is the physical part

Human Factors encompasses the physical as well as the mental and emotional.

The Man/Machine Interface

www.dialogic.com


Slide 4

PEOPLE

JUST

DON’T LISTEN!

Persistence, Memory and Time

www.dialogic.com


Slide 5


Telephony Interfaces

Vs.

Visual Interfaces


www.dialogic.com


Slide 6

Persistence

In a visual display, data remains on the display until replaced by new data.

This allows users to:– Return to a task after interruption– Review – by scanning back and forth – among several

possible menu choices– Eliminate or minimize the effects of time by scrolling freely

between the past and the present– Maintain context – even when confronted with multiple tasks

www.dialogic.com


Slide 7

Memory

The serial presentation of auditory information places heavy demands on working memory

More impactful on novice users More impactful on older users

www.dialogic.com


Slide 8

Time

Time is the enemy of the spoken user interface

-Bruce Balentine/David P. Morgan, How To Build a Speech Recognition Application

– Defeating this enemy requires repeating critical information until it “sticks”

– Yet it takes time to say things– “Hold on – I’m writing this down”– Cultural/Social issues can cause communication breakdown

• Issues of Prosody/Timing• What’s your phone Number?

– Is it 973-555-1212 or is it 9735-5-1212?

www.dialogic.com


Slide 9

Machine Output

Prompts

www.dialogic.com


Slide 10

Machine Spoken Output

Prompts – indicate it is time for user input.

Feedback – presents the application state that results from user input, allowing the user to compare original intent with final results.

Instructions – give information to the user about operating the user interface or understanding the task.

Help – offer context sensitive corrective action. Often adopts a separate mode or state aimed at coaching.

Application Data – the content or information that the user seeks or intends to modify.

www.dialogic.com


Slide 11

Silence, the Silent Killer

People will wait without feedback for six to eight seconds.– Anything longer than that and callers will think something

wrong.– Causes frustration.– Calles people to hang up.

If a processing delay or a wait in queue lasts more than six seconds, give the caller feedback.– Music, Information, Advertising.– If using tones, explain the tone or callers may think the tone

is an indication that they have been disconnected.

www.dialogic.com


Slide 12

Prompts

Prompts

www.dialogic.com


Slide 13

Action-Goal vs Goal-Action

Action-Goal– Press one for sales…

Goal-Action– For sales, press one…

Goal-Action reflects the way people think, using Action-Goal can cause confusion.

What you are saying:Press One for Sales…Press Two for Marketing…Press Three for Support.

What is heard:Press One (not heard because the user is not paying attention yet) for Sales, press two…for Marketing…press three, for support…???

www.dialogic.com


Slide 14

Please, Now and Thank-You

“Social Graces” just add to the length of the communication– For sales, please press one now…– For sales, press one…

Many phone-based interfaces are tedious because they unnecessarily put the word “please” in front of every acdtion statement on a menu (e.g., “For more information, please press 4.” (Scumacher)

www.dialogic.com


Slide 15

Anthropomorphism

The attribution of human characteristics to non-human beings

This is not the same as the system having a “personality” Experts disagree on the use of anthropomorphism

In my opinion: Avoid anthropomorphism The more “like” a person people believes the system to be

more they want to communicate with it like it is a person, but it is not a person, it is a machine

If you must personify, let the personality be a narrator or guide, not the machine

www.dialogic.com


Slide 16

Compression

The speed or tempo at which recordings are played– Should be between 135 words/min and 170 words/min– Software can be used to compress (or speed-up) playback

while maintaining the pitch of the voice– Faster may seem better, but it can cause error due to

retention issues and response mistakes…especially in older adults (Sharit, 2003)

Faster tempo can cause Perceived Enunciation Errors or Mondegreens– There’s a bathroom on the right

• There’s a bad moon on the rise– Mairzy doats and dozy doats and liddle lamzy divey

• Mares eat oats and does eat oats and little lambs eat ivy– For information and directions, press 5…

• ???

www.dialogic.com


Slide 17

Short Prompts vs Long Prompts

With dial-through, dial-ahead and/or barge in, why is this relevant.– Prevents repeating irrelevant prompts during error correction.

• “Thank you for calling XYZ, please enter your PIN”• “That was not a correct entry”• “Thank you for calling XYZ, please enter your PIN”

– Prompts can be used for “grunt detection” even when ASR is failing.

www.dialogic.com


Slide 18

Lists, Menus and User Input

Lists and Menus

www.dialogic.com


Slide 19

Hierarchy vs Skip and Scan

Hierarchy– For Sales, press 1…For Support, press 2.– There are four matches…For Jan Smith, press 1…For John Smith,

press 2…For Ken Smith press 3.– “Lakeview Terrace”, press 9…”Burn after Reading” ,press

10…”Igor”, press 11.

Skip and Scan– Sales. To select this option, press 1. For the next option, press 9.

For the previous option, press 7.– Jan Smith. To select this option, press 1. For the next option, press

9. For the previous option, press 7.– “Lakeview Terrace” To select this option, press 1. For the next

option, press 9. For the previous option, press 7.

www.dialogic.com


Slide 20

Number of Choices Per Menu

The primacy and recency effects– Designers should also consider the primacy and recency

effect that enables users to remember the first and last options most frequently. The recency effect makes the last few items presented in a list the easiest to recall. However, a short disturbance or interference can make it difficult to remember the last few items (Baddeley, 1999).

Most people can only remember 5 choice– Some can remember more, some less– More complex instructions are harder to remember– Older users have more difficulty remembering– 5 items +2, depending on user base and complexity, is a

good rule of thumb

www.dialogic.com


Slide 21

Delimiters: To # or not to #

What is that thing (#) called– Pound, Number Sign, Hash

Telling them where it is– The # Key is located at the lower right corner of your keypad.

Enter your 4 Digit PIN followed by the #?– Why required the # if you know length of the expected input?

Enter your 4 Digit PIN– What to do if they enter # anyway?

www.dialogic.com


Slide 22

Other User Inputs

Lack of Instruction, Preparation. Directional Metaphors Consistent use of keys Mnemonics Dynamic Menus Alphabetic Input

– Two Button – Key then Position– Two Button – Key then Location– Count along the key

www.dialogic.com


Slide 23

Press vs Enter

Use Press when a single digit entry is required– Implies that no Delimiter (#) is Needed– “For Sales, Press 1...”

Use Enter when a multi-digit entry is required– Doesn’t matter if it is a fixed-length entry or a variable-length

entry– “Enter your 4-digit PIN Now”– “Enter you PIN, followed by the # key”

www.dialogic.com


Slide 24

DTMF or ASR: Different or Better

DTMF: STRENGTHS– Familiarity– Ubiquity– Speed– Privacy– Efficiency– Availability– Cost

DTMF: WEAKNESSES– Auditory Only– Taxes Working Memory– Limited Input Device– Variability in Equipment

ASR: STRENGTHS– Hands Free in a Mobile World– Flexible– Adaptable– Good for Data Intensive Input

• Automated Attendant• Lists

ASR: WEAKNESSES– Cost– Difficult to Recover From Errors– Error Amplification– Regional Issues– Legally Ambiguous

www.dialogic.com


Slide 25

ASR Menus

Don’t mimic DTMF menus– “To Pay with Visa, press 1 or say one”– “To Pay with Visa, press or say 1”– “To Pay with Visa, say Visa”

How about– “What Credit Card Would You Like to Use to Pay for That”

www.dialogic.com


Slide 26

Feedback

Presents the application state that results from user input, allowing the user to compare original intent with final results:

Echoing user input for confirmation– You entered “ABC”, if this is correct, press 1, if you need to try

again press 2– You said “ABC”, is this correct?

Do not echo menu choices– For technical support press 1…

“Technical Support Menu”

Can be tedious for experienced users, the feedback can be implied in the follow up prompt

“For new product installation support, press 1, for trouble shooting an existing implementation, press 2

www.dialogic.com


Slide 27

References

How to Build a Speech Recognition Application– Balentine & Morgan, 2001

It’s Better to be a Good Machine than a Bad Person– Balentine, 2007

Increasing the Usability of Interactive Voice Response Systems: Research and Guidelines for Phone-Based Systems– Scumacher, Hardzinski & Schwarz, 1995

Skip and Scan: Cleaning up Telephone Interfaces– Resnick & Virzi, 1992

Effects of Age, Speech Rate, and Environmental Support in Using Telephone Voice Menu Systems– Sharit, Czaja, Nair, Lee, 2003

www.dialogic.com


Slide 28

Dialogic, Dialogic Pro, Brooktrout, Cantata, SnowShore, Eicon, Eicon Networks, Eiconcard, Diva, SIPcontrol, Diva ISDN, TruFax, Realblocs, Realcomm 100, NetAccess, Instant ISDN, TRXStream, Exnet, Exnet Connect, EXS, ExchangePlus VSE, SwitchKit, N20, Powering The Service-Ready Network, Vantage, Connecting People to Information, Connecting to Growth, Making Innovation Thrive and Shiva, among others as well as related logos, are either registered trademarks or trademarks of Dialogic Corporation or its subsidiaries (“Dialogic”). The names of actual companies and products mentioned herein are the trademarks of their respective owners. Dialogic encourages all users of its products to procure all necessary intellectual property licenses required to implement their concepts or applications, which licenses may vary from country to country. Dialogic may make changes to specifications, product descriptions, and plans at any time, without notice.

06/08

www.dialogic.com

USE CASE(S)Any use case(s) shown and/or described herein represent one or more examples of the various ways, scenarios or environments in which Dialogic products can be used. Such use case(s) are non-limiting and do not represent recommendations of Dialogic as to whether or how to use Dialogic products.

www.dialogic.com


Slide 29

Please do not alter the design of the

template, by changing fonts,

bullets, or design elements.

PLEASE NOTE: You may need to modify this sentence for your presentation. “© Copyright XXXX Dialogic Corporation. The XXXX should be the current year, unless the presentation includes information from pre-2008, in which case XXXX should be “XXXX-YYYY,” where YYYY is 2008 and XXXX is the earliest year of creation of content that was included in a previous version of the presentation.” For example, each “golden” likely should have a copyright date of XXXX-2008.” For example, if it’s a golden for a brand new product or no golden previously existed then it would be just 2008. .

PLEASE NOTE: Remove for presentations where an NDA is not required.

The title is 28 pt. Arial Bold

The first bullet is 24 pt. Arial Bold

The second bullet is 22 pt. Arial Bold

The third bullet is 20 pt. Arial

www.dialogic.com


Slide 30

Human Factors in Voice Interface Design Jeff Dworkin Segment Marketing Manager [email protected].

Documents

time slide

dialogic corporation

world slide

company confidential

time time

manmachine interface

analytics leadership

effects of time