Steven E. Stern (left) and John W. Mullennix studied differences in persuasiveness between artificial speech and natural language.

Is the Power of Persuasion Different between
Human and Computer-Assisted Speech?

Whether it’s a sales pitch in the supermarket, a debate in the classroom, or simply lively personal conversations, persuasion is prevalent in our daily communication scenarios.

To understand how speech- impaired individuals are affected when they need to communicate messages containing persuasive content, two University of Pittsburgh at Johnstown (UPJ) faculty colleagues investigated scenarios where listeners were exposed to normal human speech and artificial speech generated from a computer. Their findings were published recently in the Journal of Applied Psychology.

Assistive technologies are enabling people with disabilities to perform many everyday tasks while working to overcome their impairments. Such technologies can include prosthetics, mobility aids, and specialized computer applications such as text-to-speech (TTS) systems. TTS systems take typed text from a computer and produce fluent, audible, artificial speech. Since little is known about whether speech-impaired people using TTS devices are as persuasive as those who do not need such devices, Steven E. Stern, assistant professor of psychology, and John W. Mullennix, associate professor and chair of the Department of Psychology, examined the perceptions of listeners with respect to this issue.

“Our research project combines our two individual specialties—language and social influence—to fill a void in the research that had been done by others,” said Mullennix, who also directs UPJ’s Speech Perception Laboratory.

“These findings not only tell us something about how people process synthesized speech, but also give us insight into how people perceive other individuals who have speech-related disabilities,” said Stern.

Stern’s expertise in the area of human interaction with technology, combined with Mullennix’ specialization in psycholinguistics and speech technology, provides a solid approach to the study of psychology, speech, and technology.

Focusing on the specific TTS systems used as talking aids for people with hearing and speaking impairments, Mullennix and Stern, along with student Stephen Wilson, examined the persuasiveness of computer-synthesized TTS speech compared to human speech.

In the initial study, listeners were presented with a persuasive argument via computer-synthesized speech or via recorded human speech. The researchers found that human speech was perceived as soft, high-pitched, less “accented,” and lengthy, when compared to artificial speech. They also found that the human speaker was perceived as more knowledgeable and truthful, yet less powerful, than the artificial speaker. And most importantly, they found that the artificial speech was less persuasive than human speech.

In another study, listeners were told that the person delivering the speech was either speech-impaired or not speech-impaired. Two versions of a persuasive appeal were presented via videotape, one using normal speech, the other using computer-synthesized speech. Listeners’ reactions to the speech were measured in terms of qualities of the speech (e.g., loudness, accent, nasal), perceptions of the message (stimulating, vague, complex), and perceptions of the speaker (competent, assertive, timid).

The results showed that human speech was perceived more favorably and was more persuasive than computer-synthesized speech. However, when the individual uttering a speech was perceived as speech disabled, his or her use of the synthetic speech system was viewed more positively than a person perceived as NOT speech disabled. This finding suggests that information about the person using a TTS system affects the perceptions and attitudes the listener has towards the message that the speaker is imparting.

Ultimately, the studies found that a person with disabilities using a TTS system is at risk of being less persuasive than a nondisabled person. However, this tendency may be mitigated by knowledge the listener possesses about the user. Thus, the disadvantage for the speaking-impaired person is diffused when the observer realizes that the person’s need is real. The findings suggest that people tend to be more forgiving when such enabling technology is seen as necessary.

The researchers are encouraged by the results of this research. Although such technology appears to be viewed negatively, people generally impart little bias to the speech disabled in their usage of TTS. Therefore, speech disabled people using TTS systems to communicate with others should not be perceived as any less persuasive in their typical encounters.

Stern and Mullennix hope their research project will prompt further interest in how nondisabled people interact with people with disabilities who need to use assistive speech technologies, and that businesses and agencies that work with and accommodate people with disabilities can find value in the findings that have been the focus of this collegial research.

“We still feel that there needs to be more work done to assess how people perceive disabled users of this technology in more realistic settings,” said Stern.

The researchers also believe that their work opens the door to other uses.

“This work also is important for understanding a side to computer-mediated communication outside of e-mail and the Internet,” said Mullennix. • KM