Today, traditional providers of news are subject to training as well as social and professional pressures to provide accurate and unbiased information, giving people a sense that what they see is credible. Conversely, the Internet's free and unregulated flow of information and information providers creates many possible hazards to those who seek and trust online information. In many instances those accessing information are unaware of: (1) who authored the material, (2) when the information was last updated, and (3)whether the information is accurate. As the Internet's information seekers and providers continue to increase, it becomes important that researchers gain an understanding of how this information is being perceived. Therefore, this study assessed two key factors concerning how people determine the credibility of online information. The relationship between topic knowledge, source expertise, and perceived credibility of the message was investigated. In this evaluation, the level of source expertise (i.e., high, moderate, and low), and content knowledge (i.e., an unknown topic and a known topic) were manipulated.
Assessing Credibility on the Internet
According to the Pew Research Center (2001), approximately 104 million American adults have access to the Internet. Of these, there are slightly more females (51%) on the Internet than males (49%), of both high and low socioeconomic status. As the Internet's audience continues to grow and subsequently mirror the general population, understanding how people use it to obtain medical information becomes more important to both users and providers.
When considering the Internet, one distinct feature pertaining to the flow of online information must be understood; unlike traditional media, the Internet has no government or ethical regulations controlling the majority of its available content. This unregulated flow of information presents a new problem to those seeking information, as more credible sources become harder to distinguish from less credible sources (Andie, 1997). Moreover, without knowing the exact URL of a given site, the amount of information offered through keyword searches can make finding a predetermined site difficult as well as increase the likelihood of encountering sites containing false information (Andie, 1997).
Presently, the medical community has become increasingly aware of and concerned about the credibility of health information available on the Internet (The Pew Research Center, 2000; Wright, 1998). A growing number of patients are seeking medical advice everywhere from chat rooms to personal Web pages (Donald, Lindenberg & Humphreys, 1998; Dudley et. al., 1996). In an early attempt to assess medically-driven online activity, Donald et al. (1998) evaluated two medically-oriented Internet sites, MEDLINE and Grateful Med. During a yearlong evaluation, 75 million searches occurred within these sites. To gauge who was seeking information, a voluntary two-item general questionnaire was administered for one week; it indicated that 30 percent of respondents were either students or from the general public. More recently, the Pew Research Center (2000) reported that 55 percent of Internet users access the Internet in search of medical information. Of these, 21 million indicated that they had been influenced by health information they read online. That is, of those influenced by online medical information, 70 percent said that their decision to treat an illness or condition was influenced by online sources, while 28 percent reported that online information influenced their decision about whether or not they should see a doctor.
These data, accompanied by previously mentioned increases in online use, suggest a large number of people use the Internet to gather medical information. Moreover, this activity brings several issues to the foreground: (1) Who are the people using the Internet for health information? (2) Can the information obtained online be trusted? And finally, (3) How do people seeking medical information online determine what sources are credible?
Addressing these issues, Culver, Gerr, and Frumkin (1997) evaluated messages retrieved from a medical chatroom discussion group. Of the 1,658 messages evaluated, people without professional experience authored 89 percent, and one-third of the messages were found to be inconsistent with conventional medical practices. Also analyzing chatroom content, Feldman (1998) documented the leader of a chat room who pretended to have cancer in order to gain attention. In that case, after gaining information by viewing others' discussions, she began to pass the information off as her own.
The medical community has tried to locate and rate Web-based health information sites. However, it has achieved little success. (Donald et al., 1998). One of the main obstacles confronting this process is the nature of Web publishing. Donald et al., (1998) note that many documents on the Web “lack basic information about the origin, authorship, or age of the material they provided” (p.1304). Moreover, when considering the ease of creating and changing a Web page, together with the very large number of Web publishers, the task of monitoring all medical information becomes insurmountable. Thus, the task of assessing the credibility of information obtained online rests with the user. In order to understand how people perceive credibility on the Internet, it is first essential to understand how more traditional media research (i.e., television and print) has explored the issue of credibility.
Early credibility research on media can be traced back to Hovland & Weiss' (1951) research on communication and persuasion. They found that the “trustworthiness” of a source significantly affected acceptance of the message and changes in opinion. Significantly related to trustworthiness were reactions to “fairness” of presentation and “justifiability” of the conclusions. Since these original studies, many variables have been examined to assess source credibility. (Adoni, Cohen, & Mane, 1984; Gaziano & McGrath, 1986; Meyer, 1988; Slater & Rouner, 1996). Perceived expertise, bias, fairness, truthfulness, accuracy, amount of use, depth or completeness of message, prior knowledge and message quality have all emerged as components of credibility.
Recognizing the large number of indicators and dimensions used to assess the construct of credibility, West (1994) validated two frequently used scales of credibility, by Gaziano and McGrath (1986) and Meyer (1988). The five dimensions created in the Meyer (1988) study were found to be: “…valid and reliable measures of credibility; they consisted of fairness, bias, depth, accuracy, and trustworthiness.” (West, 1994, alpha = .92, p. 163). These five indicators, as well as 12 others, were listed in the Gaziano and McGrath (1986) scale. When evaluating which items to include, it must be determined whether credibility of the source or of the stimuli (content) is being measured, and whether credibility should be measured from the receiver's perspective or from the source's attributes. (Rubin, Palmgreen, & Sypher, 1994). That being said, this study identified items related to the content from the receiver's perspective.
While these as well as other dimensions have been found to be significant components in understanding perceived credibility, there are other factors that have helped to gain understanding of how people perceive credibility, such as age, education, amount of use, reliance and medium (Greenberg, 1966; Johnson & Kaye, 1998; Wanta & Hu, 1994). Although earlier studies had indicated that television was a more credible information source than newspapapers for most people, a more recent evaluation of media credibility indicated that people consider information obtained online to be as credible as television, radio, and magazine information, but not as credible as information in newspapers (Flanigan & Metzger, 2000).
Relatively speaking, the Internet has only recently become part of mainstream America. Thus, it is important to begin evaluating perceived credibility of the Internet by incorporating Internet-specific characteristics. To date, research on Internet credibility has focused on four areas: media comparisons, demographics, medium reliance, and message characteristics. For example, Brady (1996) and Johnson & Kaye (1998) found information about political candidates on the Web was perceived to be just as credible as information on television.
Assessing demographics, Johnson and Kaye (1998) found gender the only variable significantly correlated with perceived credibility. Females found the Web more credible and trustworthy than males. Although not significant, age and education were negatively correlated with perceived credibility of online newspapers, online candidate literature, and politically-oriented sites. Results also indicated that levels of medium reliance and use were positively correlated with perceived credibility. Reliance on the Web for political information was a stronger predictor than use of perceived credibility.
In an effort to evaluate the dynamics of online source and content, researchers have attempted to assess whether or not individuals use heuristic cues when source credibility is limited or questioned within online chatrooms (Franke, 1996; Slater & Rouner, 1996). Results indicate that most discussions are too brief or unusual to be considered usable dialogue, which suggests that actual Web pages will make a more profitable platform from which to evaluate perceptions of message credibility.
More recently, Rieh & Belkin (1998) identified criteria used when evaluating online information. Specifically, using in-depth interviews with 14 scholars, they assessed perceptions of information quality and authority of information. For example, they found that: (1) institutional sites were seen as more credible than individual sites, and (2) accuracy of content was used to assess online information. Respondents used knowledge of citations within the content and the functionality of hyperlinks as cues to evaluate the information. Similar variables were indicated in a report focusing on assessing the quality of online health information. (Agency for Health Care Policy and Research, 1999). In addition to source and link accuracy, they also recommend that users consider peer evaluation, navigability, and feedback options (i.e., email, chat room, etc.).
While these Internet-based studies have helped to bridge the gap between traditional and online media research by offering insight into perception of the Internet, researchers have not addressed the issue of how people using the Internet to gather information assess the credibility of content provided. Therefore, an area within the credibility literature that needs to be explored within an online context is content type.
Persuasion and Context
Persuasion scholars have shown how various elements of messages such as language intensity, style, attractiveness, and quality all effect message perceptions (Adoni et al., 1984; Chartprasert, 1993; Hamilton, 1998; Petty & Cacioppo, 1979; Slater & Rouner, 1996). In addition, knowledge of content has been identified as one of the more significant factors impacting heuristic cues used when evaluating a message (Eagly & Chaiken, 1993; O'Keefe, 1990). Research on persuasive messages has found that knowledge mediates receivers' ability to evaluate messages critically. (Eagly & Chaiken, 1993). When receivers are highly knowledgeable, their ability and motivation to process are greater, which heightens their critical interpretation.
In this context, the persuasion literature suggests that respondents' processing heuristic cues establish expectancies about the message validity. For example, Cozzens and Contractor (1987) evaluated how people evaluate message content when the subject matter was familiar. They demonstrated that active audience members evaluating a mediated message used outside understanding (i.e., knowledge from a personal source) of the issue to evaluate the believability of the message rather than of the source. This suggests individuals assess the content of a message based on extrinsic information
In addition to prior knowledge of content, limited knowledge of source competence as well as low involvement with the subject matter also cause respondents to seek message-inherent heuristic cues (i.e., presentation quality, language intensity or style, attractiveness, and subjectivity) apart from the source to evaluate the message. Similar to content knowledge, cues embedded in the message become characteristics with which to evaluate the message validity. For example, as personal relevance increases or where knowledge of source credibility is limited, respondents become more motivated to process issue-relevant content (Eagly & Chaiken, 1993; Petty & Cacioppo, 1988). This factor, together with the aforementioned findings, underscores the importance of understanding that while the source of a message is a commonly used attribute to assess the perceptions of credibility, other content-driven variables can also affect message perceptions and thus should be considered.
This study builds on these findings by varying source expertise and content knowledge within an online environment. To gain needed understanding of how Internet users perceive message credibility, this study, utilizing items found in Gaziano & McGrath, (1986), Meyer, (1988) and Johnson and Kaye (1999), evaluated perceived credibility of online information. As suggested, people publishing information online vary in expertise, and past credibility research suggests that expertise is a main predictor of perceived credibility (Newhagen & Nass, 1988; Slater & Rouner, 1996). Therefore, using types of sources commonly found on the Internet, this study manipulated expertise into three levels. Specifically, a linear expertise continuum consisting of three levels of expertise was established. Level 1 consists of a highly rated source; Level 2 consists of a moderately rated source; and Level 3 consists of a source considered not at all an expert.
Many studies also suggest that content type can mediate source effects (Austin & Dong, 1994; Brady, 1996; Slater & Rouner, 1998). Therefore, borrowing from previous research (Cozzens & Contractor, 1987; Rieh & Belkin, 1998), this study manipulated: (1) information about one unknown topic; and (2) information about one known topic. The content-type manipulation helps to provide an understanding of the relative importance of Web page source and message content. Research suggests that when participants are knowledgeable about the subject matter, they are more apt to attend to the message, rather than to the source (Adoni et. al., 1984).
It therefore seems logical that when people have little knowledge of online content, they will perceive information attributed to a highly expert source as more credible than information from a less expert source. Likewise, information from a moderately expert source should be perceived as more credible than a source low in expertise. When people are highly knowledgeable about message content, the effects of source expertise will be mediated (i.e., attenuated). From this, it is hypothesized that:
H1 There will be a significant interaction between source expertise and knowledge of content. That is, for the topic where subject matter knowledge is low, information attributed to moderate and low expert sources will be perceived as less credible than identical information attributed to a highly expert source. For the topic where subject matter knowledge is high, source expertise will have no significant effect on perceived credibility.