Interactive Voice Response IVR Systems and Older Adults: Lessons Learned

1
Interactive Voice Response (IVR) Systems and
Older Adults Lessons Learned

• Michelle McNulty
• Ellen Connor Mangan
• Fidelity Investments
• Boston, MA

Currently employed by The Mathworks
2
Introduction

• Interactive Voice Response (IVR) systems have
  become more complex and more prevalent.
• Information can be accessed and processed over
  the phone through touch tone and speech user
  interfaces.

3
Introduction

• The speech component allows customers to use
  verbal commands to obtain general or personal
  (account) information and conduct transactions.
• Providing an intuitive user interface is key for
  a successful user experience.

4
Fidelitys Phone Stats

• Fidelitys phone systems receive a combined
  average of 282,000 phone calls per day.
• Forty-two percent of those calls are handled by
  an automated system.

5
Fidelitys Phone Stats

• Usage statistics for the systems report that
• 79 of the callers are 50 or older
• 51 of the callers are 60
• Less than 5 of the callers are under the age of
  40
• Because older users may be less comfortable with
  the web, the voice channel remains an important
  part of their overall user experience.

6
Background

• Fidelity has conducted a great deal of research
  into older adults usage of the web.
• We observed many age-dependent behaviors that
  affected web usage.
• Would we see similar results if we looked at our
  IVR system?

7
Our Usability Studies

• We have conducted over 125 IVR-based usability
  and research sessions over the past 5 years, many
  with older adults.
• The studies included several typical usability
  sessions (approx. 8-10 users) and one larger
  research study that specifically focused on older
  users.

8
Our Usability Studies

• The systems tested were designed to
• Find account and personal information
• Conduct transactions
• Report technical problems
• Check voicemail
• Set up a voice ID
• Most systems used both speech recognition and
  touch tone interfaces.

9
Research Study Results
Task success rate declined as a function of age.
10
Task Completion Rate by Age Group

• (p .037, n27)

11
Designing for the Older User

• Because of physical and cognitive changes that
  happen with age, we expected to see performance
  decline with age.
• Reduced working memory capacity
• Hearing
• Motor dexterity
• Understanding the changes we undergo as we age
  should influence the design of systems that have
  a significant population of older users.

12
Designing for the Older User

• Older users bring different abilities and
  expectations to phone-based interactions than
  younger users do.
• Until the middle of 2004, 96 of users interacted
  primarily through touchtone.
• Speech recognition and the primary use of voice
  commands is relatively new.
• What issues would older users encounter?
• What benefits would older users discover?

13
Lessons Learned

• We compiled the following lessons learned from
  the variety of usability studies we conducted
  over the past few years with older adults.
• Our recommendations are based on our past
  experience and other available research.

14
Dont set false expectations

• Unrealistic expectations of systems abilities
  (Expecting a Star Trek experience).
• Users knew they werent talking to people, but
  many older adults tried to interact with the
  system as if it were human.
• The conversational and easy-going nature of
  some interfaces may set false expectations.
• Some users described certain system responses,
  such as Got it, as colloquial and not
  professional.
• Younger users expressed annoyance at the system
  attempting to sound sorry.

15
Be Consistent

• Consistency is very important.
• Sudden auditory changes of the automated voice
  might signal an error or incorrectly indicate
  that users have been transferred somewhere else.
• Menu structure and controls should be consistent
  throughout the interaction. Dont set up the
  ability to use voice commands in one part of the
  system and then disable voice commands in other
  parts of the system.

16
Pacing of System

• Pace of system was often too fast for older
  users.
• Older users tend to be slower at information
  processing.
• Providing them just enough time between spoken
  menu items to echo each choice sub-vocally
  appears to improve their success with IVR
  Systems.
• However, some younger users actually preferred
  menu items spoken at a more rapid pace (less time
  between items) as they could get to the
  information they wanted faster.

17
Pacing of System

• Pace of system was often too fast for older
  users.
• If possible, consider implementing a system that
  recognizes the age of the user and leverages that
  to determine the amount of space between items
  and the speed at which the menu prompts are
  spoken.
• If that is not possible, we recommend following
  the option that would help the greatest number of
  users, i.e. the largest user group.

18
Avoid Information Overload

• Too many menu choices long messages
• Reduced working memory capacity in older users
  made this more problematic for them than younger
  users.
• When presented with a long menu, we observed many
  older users simply choosing the last item in the
  list.
• Lengthy messages can also overload memory. We
  observed a a few older users opting out to a
  representative possibly because it took too long
  to get to the menu prompts.

19
Avoid Jargon

• Usage of jargon, obscure terminology
• This affected all users, but especially older
  users and those with lower levels of financial
  literacy.
• If users are unsure what a menu prompt means,
  they will be more likely to just make a random
  choice or attempt to reach a representative
  quickly.
• Know your users and avoid using jargon-y terms
  that they may not understand.
• Using familiar words ones that refer to
  concrete objects has been shown to improve
  performance.

20
Politeness

• Older users were overly polite.
• Older users were often hesitant and had trouble
  interrupting the system.
• They often said please and thank you.
• They spent much longer on the phone at the end of
  the call than younger users because they waited
  for an obvious out vs. just hanging up.
• Allow generous barge-in.
• The system should ignore various pleasantries
  such as please and thank you.
• Inform users at various points in the interaction
  that they can hang up when theyre finished.

21
Include Some Touchtone Options

• Many users prefer using the touchtone options to
  speaking their choices.
• Always offer a touchtone option for entering
  sensitive information, like SSN and PIN.
• Touchtone options serve as an external memory
  device, offering a physical indication of the
  available choices (think of a finger hovering
  over buttons). This seemed especially helpful to
  older users.

22
Include Some Touchtone Options

• Use touchtone options strategically when
  transitioning users into a new technology
• In the initial rollout, offer both touchtone and
  voice options in the first menu.
• Touchtone options should always be available as a
  backup.
• Accents, environments, privacy concerns.

23
Frequency of Sound

• Because higher-frequency sounds are the first to
  go with progressive hearing loss1, use a
  lower-pitched voice.
• Peak hearing sensitivity is around 3,000 4,000
  Hz.
• Lower frequencies are less susceptible to
  masking effects

1 National Academy on an Aging Society Hearing
Loss http//www.agingsociety.org/agingsociety/pd
f/hearing.pdf
24
Positive Reaction to Voice Recognition

• Many seniors were very enthusiastic about the
  voice recognition.
• Hands-free interaction helps users
• With physical impairments such as arthritis
• Fine motor-control problems
• On cell and cordless phone
• Words are much easier! comment from a user
  with severe arthritis.

25
Conclusion

• Many of the usability issues we observed affected
  both older and younger people, but in general
  older users had more trouble recovering from
  those errors.
• Many of the usability recommendations given here
  would benefit all users if implemented.
• Understanding issues that result from the aging
  process help not only older users but lead to a
  more universally accessible design.

26
Conclusion

• The key is to know and design for your user.
• If the majority of users is older make sure your
  system accommodates their requirements.
• Set expectations. Warn customers of big changes
  ahead of time and/or in the system itself.

27
Contact Information

• Michelle McNulty
• Michelle.McNulty_at_fmr.com
• Ellen Connor Mangan
• Ellen.Mangan_at_mathworks.com

28
Resources

• Andrade, J. (2001). An introduction to working
  memory. In J. Andrade (Ed.), Working memory in
  perspective (Chapter 1, pp. 3 30). Hove, UK
  Psychology Press
• Banbury, S., Macken, W, Tremblay, S., and Jones,
  D. (2001) Auditory Distraction and Short-Term
  Memory Phenomena. From Human Factors. Volume 43,
  Issue 1. 2001
• Gupta, P. and MacWhinney, B. (1993) Is the
  Phonological Loop Articulatory or Auditory?
  Proceedings of the Fifteenth Annual Conference of
  the Cognitive Science Society. Lawrence Earlbaum
  Hillsdale, NJ. 1993.
• Noonan, Tim Building User-Friendly Voice
  Systems, 1998, Downloaded from http//www.timnoona
  n.com.au/ivrpap98.htm on April 16th, 2006
• Stolzfus, E., Hasher, L., and Zacks, R. (1996)
  Working Memory and Aging Current Status of the
  Inhibitory View. From Working Memory and Human
  Cognition. Oxford University Press, New York.
  1996.