Support seeker expectations,support message quality,and supportive interaction processes and outcomes: The case of the comforting computer program revisited

Abstract

Developing a complete understanding of supportive communication in personal relationships requires considering the role of support seekers. We examined the influence of seeker expectations and verbal person centeredness (VPC) on the nature and outcomes of supportive interactions. Participants discussed a personal problem during an interaction with a fictional computer program called “ListenerBot.” Participants’ expectations about the helpfulness of ListenerBot and the quality of ListenerBot’s feedback in the form of VPC were manipulated. Participants in the helpful prime condition wrote more words during the interaction and evaluated ListenerBot more favorably than participants in the unhelpful prime condition. Relative to participants who received low VPC feedback, participants who received high VPC feedback evaluated ListenerBot more positively and experienced a greater reduction in emotional distress. Tests for indirect effects showed that VPC and the expectations prime contributed to participants’ reduction in emotional distress through influencing their evaluations of ListenerBot.

Keywords

Behavioral confirmation emotional distress human–computer interaction social support support seeking verbal person centeredness

Social support is a critical coping resource that has garnered the attention of scholars from numerous disciplines (Cohen, Underwood, & Gottlieb, 2000; Goldsmith, 2004; Thoits, 2011). One prominent strand of research has focused on supportive communication and examined the qualities of more and less effective support messages (for a review, see MacGeorge, Feng, & Burleson, 2011). Verbal person centeredness (VPC) is an important element of emotional support messages, involving the degree to which a message validates a recipient’s distress and helps the recipient to better understand his or her feelings (Burleson, 1987). Researchers have routinely documented the benefits of high VPC messages for coping with emotional distress (for a review, see High & Dillard, 2012). Indeed, the salutary effects of high VPC messages have even been observed in the context of standardized support messages delivered by a computer program (Rains, Pavlich, Tsetsi, & Akers, in press).

In the present study, we extend scholarship on supportive communication in personal relationships by further examining the scope of VPC effects and spotlighting the role of support seekers. Although support seekers have frequently been relegated to the role of passive message recipients in research on message features like VPC (for exceptions, see Barbee & Cunningham, 1995; High & Scharp, 2015), their perceptions and behavior likely influence the nature and outcomes of supportive interactions (Buehler, Peterson, & High, 2018; Oh & LaRose, 2016; Youngvorst & High, 2018). We focus specifically on the implications of seekers’ expectations about the helpfulness of a support provider. Research on behavioral confirmation (Snyder & Swann, 1978; Snyder, Tanke, & Berscheid, 1977) suggests that the expectations of support seekers should shape their perceptions, behavior, and the outcomes they ultimately experience from supportive interactions. We report the results of an experiment in which participants interacted with what appeared to be a computer program designed to listen to people’s problems. Support seekers’ expectations about the potential helpfulness of the program and the level of VPC in the program’s responses were manipulated to examine their effects on seekers’ efforts in soliciting support, perceptions of the program, and change in emotional distress. This project advances scholarship on social support in personal relationships by further documenting the effects of VPC along with highlighting the role of support seekers and exploring how their expectations prior to a discussion can shape the nature and outcomes of supportive interactions.

Support seekers and supportive communication

Supportive communication involves verbal and nonverbal behavior used to assist people in need of aid (MacGeorge et al., 2011). Although the message qualities of advice (MacGeorge, Feng, & Guntzviller, 2016) and esteem support (Holmstrom, Clare, & Russell, 2014) have received attention, much of the work on this topic has focused on examining the effects of VPC. As a quality of emotional support messages, VPC can be classified on a hierarchy that has three broad levels, each of which contains three more specific subcategories (Burleson, 2008). The lowest levels of VPC involve criticizing and denying the feelings of support seekers, whereas messages containing moderate levels of VPC implicitly recognize support seekers’ feelings. Messages containing high levels of VPC legitimize and explicitly engage the feelings of support seekers.

The theory of conversationally induced reappraisals (Burleson & Goldsmith, 1998) explains the effects of VPC on support seekers’ coping efforts. Drawing from Lazarus’s work on appraisal in stress and emotions (Lazarus, 1999; Lazarus & Folkman, 1984), this theory suggests that relief from emotional distress occurs via reappraisal as people come to see their distress-inducing circumstances or their responses in a more adaptive way. Messages containing high levels of VPC can play a critical role in fostering reappraisal (Burleson & Goldsmith, 1998; Jones & Wirtz, 2006). High VPC feedback can encourage people to discuss their feelings and engage in sensemaking that leads to reappraisal and, ultimately, a reduction in emotional distress.

Researchers examining the effects of VPC have routinely demonstrated the benefits of high VPC messages (Samter & MacGeorge, 2016). High and Dillard (2012) conducted a meta-analysis of 23 studies and found a robust relationship between the level of VPC in support messages and positive outcomes, although noteworthy sex differences in VPC effects have been reported in several studies (e.g., Burleson, Holmstrom, & Gilstrap, 2005; High & Solomon, 2014). Consistent with the theory of conversationally induced reappraisals (Burleson & Goldsmith, 1998), researchers have also shown that high VPC messages can encourage support seekers to discuss their emotions and use language that reflects sensemaking (Bodie, 2011; Jones & Wirtz, 2006).

Although VPC research has made important contributions to our understanding of supportive communication, this body of scholarship may be critiqued on a few grounds. Perhaps most notably, the vast majority of studies examining the effects of VPC have involved asking participants to evaluate this message feature in the context of hypothetical scenarios (High & Dillard, 2012). Despite the merits of scenario-based research, relatively few studies—only two were identified in High and Dillard’s (2012) meta-analysis—have attempted to demonstrate whether or not the benefits of high VPC messages extend to support seekers discussing their own personal problems. Similarly needed is research further exploring the theory of conversationally induced reappraisals (Burleson & Goldsmith, 1998) by examining the effects of VPC on the way support seekers talk about their problems.

In the present study, we attempt to address these issues by employing a computer program to provide high and low VPC feedback in real time as support seekers discuss a personal problem. Researchers studying human–computer interaction have known for some time that people mindlessly apply rules and expectations for interaction with humans to interactions with computers (Nass & Moon, 2000; Reeves & Nass, 1996). These social responses have been shown to apply to things like politeness behaviors, expectations about personality, and even supportive communication. Participants in two different studies, for example, responded favorably when a computer program expressed empathy (Brave, Nass, & Hutchinson, 2005; Liu & Sundar, 2018).

Although the connection may not be obvious, research examining human–computer interaction has the potential to generate valuable insights about basic interactional processes in personal relationships. In knowing that people mindlessly apply rules and expectations for human behavior in their interactions with machines (Nass & Moon, 2000; Reeves & Nass, 1996), studies considering human–computer interaction can inform us about those underlying rules and expectations. Most relevant to this study, the benefits of high VPC support messages have been documented in research exploring human–computer interaction. In two experiments, participants discussed a personal problem during an interaction with what appeared to be a computer program that provided standardized feedback containing high or low levels of VPC (Rains et al., in press). Participants in both experiments who received high VPC feedback experienced a significantly greater reduction in emotional distress than participants who received low VPC feedback. VPC, however, did not influence participants’ talk about their emotions. We extend this line of work in the present project by examining the effects of VPC and expectations of provider helpfulness.

Prior to addressing provider helpfulness, we consider the effects of VPC on three sets of outcomes. First, the level of VPC in a support message should influence seekers’ evaluations of support providers. The face-threatening nature of low VPC feedback, which involves criticizing or denying support seekers’ feelings (Burleson, 2008), should lead seekers to evaluate providers less positively than when they receive high VPC feedback. We also expect that high VPC feedback will lead support seekers to engage in greater discussion of their problems than low VPC feedback. This prediction stems directly from the theory of conversationally induced reappraisals (Burleson & Goldsmith, 1998). The validation and efforts to foster understanding that define high VPC feedback should make support seekers more motivated to engage their problems and discuss those problems more extensively. Although the engagement spurred by support containing high levels of VPC should extend to seekers’ efforts to talk about their emotions and use language reflecting sensemaking (Burleson & Goldsmith, 1998), the results from research testing these outcomes have been mixed (Bodie, 2011; Jones & Wirtz, 2006; Rains et al., in press). These inconsistent findings underscore the importance of further examining the influence of VPC on the ways in which people talk about their problems. Finally, given prior research demonstrating the benefits of messages containing high levels of VPC (High & Dillard, 2012; Samter & MacGeorge, 2016), we expect that high VPC feedback will lead support seekers to experience a greater reduction in emotional distress than low VPC feedback. Taken as a whole, this set of predictions serves to replicate previous research examining the effects of high and low VPC messages delivered by a computer program (Rains et al., in press) and further evaluate the mechanisms of VPC (Bodie, 2011; Jones & Wirtz, 2006).
Hypothesis 1 (H1): Relative to support seekers who receive low VPC feedback, seekers who receive high VPC feedback will (a) evaluate the support provider more positively, (b) engage in greater discussion of their problems indexed by total words spoken and use of language reflecting emotions and sensemaking, and (c) experience a greater reduction in emotional distress.

The role of support seeker expectations

Although message qualities like VPC are important to consider in supportive communication research, so is the role of support seekers as active participants in supportive interactions (Buehler et al., 2018; High & Scharp, 2015; Oh & LaRose, 2016). The nature of supportive communication and the outcomes experienced by support seekers are at least partially contingent upon their perceptions and behavior in soliciting aid (Barbee & Cunningham, 1995; Youngvorst & High, 2018). One important factor involves the expectations that support seekers have about support providers and supportive interactions. Researchers have shown that support seekers’ expectations about the types of feedback typically offered by trained help providers (e.g., information, advice, emotional support, etc.) can influence their evaluations of a specific provider’s efforts to give aid (Jay, Afifi, & Samter, 2000). Moreover, support deficits stem from the failure of support providers to meet seekers’ expectations and needs (McDowell, Occhipinti, Ferguson, Dunn, & Chambers, 2010; Seiger & Wiese, 2011). Taken together, previous research indicates that expectations play an important role in supportive interactions and suggest that they could influence support seekers’ perceptions and behavior in acquiring assistance and, ultimately, the success of their coping efforts.

Although there are multiple models that offer insights about the effects of expectations for others’ behavior (e.g., expectancy violations theory; Burgoon & Hale, 1988), behavioral confirmation (Snyder & Swann, 1978; Snyder et al., 1977) is uniquely suited to explain how expectations influence the perceptions and actions of support seekers in soliciting aid. Behavioral confirmation is akin to a self-fulfilling prophecy (Merton, 1948). It generally involves the process through which our expectations about an interaction partner lead us to behave in ways that cause our partner’s behavior to conform to our initial expectations. In expecting that an interaction partner will be discourteous, for example, we will behave rudely and provoke an impolite response from our partner, which serves to confirm our initial expectation.

The process of behavioral confirmation has two major parts (Snyder & Swann, 1978). The first, which is isolated in this project, occurs when our expectations about an interaction partner lead us to engage in behavior that is consistent with those beliefs. The expectation that a partner is attractive rather than unattractive, for example, has been shown to lead people to be more sociable in telephone or computer-mediated interactions where the physical appearance of their partner was withheld (Snyder et al., 1977; Tong & Walther, 2015). The second part occurs as our behavior toward our partner, based on our initial expectations, shapes our partner’s response. Our interaction partner comes to behave in ways that are consistent with our initial expectations. The negative cognitions involved in loneliness, for instance, can lead people to elicit social rejection from others (Cacioppo & Hawkley, 2009).

Behavioral confirmation (Snyder & Swann, 1978; Snyder et al., 1977) offers a valuable perspective for understanding the implications of support seekers’ expectations—particularly those expectations about the likelihood of a support provider to be helpful. It suggests that the expectations support seekers have should influence their perceptions and behavior during supportive interactions as well as the outcomes they ultimately experience. The expectations seekers have about the potential helpfulness of a provider should color their perceptions of the interaction. Support seekers should be motivated to look for and attend to cues from the provider that correspond to their initial expectations. Behavioral confirmation should lead support seekers to evaluate the provider in a manner consistent with their initial expectations. Beyond evaluations of the provider, the effects of behavioral confirmation should extend to seekers’ behavior during the interaction and the outcomes they experience. When seekers expect a support provider to be helpful, they should be more motivated to invest in the interaction. Support seekers should engage in behaviors like talking extensively about their thoughts and feelings related to the problem because they anticipate that the interaction will be helpful. Finally, theorizing about behavioral confirmation suggests that seekers’ expectations should directly influence the outcomes they experience from the interaction. Expectations should function like a placebo effect in which the belief that one is (un)likely to yield assistance is alone sufficient to impact one’s emotional distress. We summarize this set of predictions in the following hypothesis.
Hypothesis 2 (H2): Relative to support seekers who receive a prime indicating that the provider will be unhelpful, seekers who receive a prime indicating that the provider will be helpful will (a) evaluate the provider more positively, (b) engage in greater discussion of their problems indexed by total words spoken and use of language reflecting emotions and sensemaking, and (c) experience a greater reduction in emotional distress.

The preceding arguments imply a series of indirect effects. The helpfulness expectations prime and VPC lead to a reduction in support seekers’ emotional distress by encouraging seekers to view support providers more positively and engage in more extensive discussion of their problems. Viewing support providers more positively should make their feedback more meaningful, and more extensive discussion of one’s thoughts and feelings should better facilitate reappraisal (Burleson & Goldsmith, 1998; Jones & Wirtz, 2006). Accordingly, seekers’ evaluations of providers and discussion extensiveness should mediate the effects of helpfulness expectations and VPC on the change in emotional distress experienced by seekers.
Hypothesis 3 (H3): (a) VPC and (b) the helpfulness expectations prime indirectly influence support seekers’ change in emotional distress through encouraging seekers to engage in more extensive discussion of their problem.

Hypothesis 4 (H4): (a) VPC and (b) the helpfulness expectations prime indirectly influence support seekers’ change in emotional distress through encouraging seekers to evaluate the support provider more positively.

Method

Participants

A total of 266 participants consisting of undergraduate students at a large university in the U.S. completed the experiment. The data from 18 participants were omitted because they had previously completed the study, and an additional 13 participants were removed because they failed the attention check, did not follow instructions, or due to technical issues. The final sample consisted of 234 participants. On average, participants were 20.97 years old (SD = 2.07) and more likely to be female (n = 161; 68.8%). Almost three-quarters of the sample self-identified as White (n = 169; 72.2%).

Design

A 3 (expectations prime: helpful/unhelpful/no prime) × 2 (support quality: high VPC feedback/low VPC feedback) between-participants design was used in this study. The design was fully crossed.

Procedure

Data collection took place at individual computer stations in a research lab, and informed consent was secured from participants prior to beginning the study. Participants were informed that they would be interacting with a computer program called “ListenerBot” that was being developed by the researchers to listen to people’s problems. ListenerBot offered a means to manipulate the independent variables during a real-time supportive interaction. Rather than asking participants to respond to a hypothetical scenario as is common in VPC research (High & Dillard, 2012), the approach used in this project made it possible for participants to discuss their own personal problems in the context of an interaction in which they received supportive feedback.

Prior to interacting with ListenerBot, participants first identified a current problem they were willing to discuss related to money (n = 40), work (n = 21), academics (n = 102), or personal relationships (n = 71). Participants reported their level of emotional distress about their problem and then were exposed to the helpfulness expectation prime. All participants were told that ListenerBot was in a testing phase. Participants were randomly assigned to receive additional information about the effectiveness of ListenerBot.

In the unhelpful prime condition, participants were informed that preliminary testing had been unsuccessful over the course of approximately 500 trials. ListenerBot had been rated very unhelpful 95% of the time, and users almost never reported feeling better after sharing their problems. Participants in the helpful prime condition were told parallel information with the trials presented as successful and ListenerBot deemed helpful. To further emphasize the expectations prime, participants were provided anonymous feedback purportedly from the last five users. All five comments were uniformly positive in the helpful prime condition (e.g., “ListenerBot was totally worth it…great feedback,” “I got a TON out of talking to ListenerBot”) and negative in the unhelpful prime condition (e.g., “ListenerBot was a total waste…terrible feedback,” “I got NOTHING useful out of talking to ListenerBot”). A third no prime condition was included in which participants did not receive any information about previous tests of ListenerBot’s efficacy. All participants then completed a manipulation check question for the helpfulness prime along with two other questions included to camouflage the prime manipulation check.

Participants were next directed to a separate web page where they interacted with ListenerBot. To reinforce the idea that participants were interacting with a computer program, ListenerBot was designed to resemble an early MS DOS-based program consisting of only a sparse, black background and green text. A series of videos were embedded to animate ListenerBot. Words became visible one letter at a time to give the appearance that they were being generated by the ListenerBot program. The interaction proceeded in a question-and-answer format. ListenerBot posed a question to the support seeker, which the seeker then answered. Depending on the condition, ListenerBot responded with high or low VPC feedback and then asked another question. This process occurred for five total questions–answer–response turns.

VPC was manipulated in ListenerBot’s feedback to participants’ answers. High and low VPC messages originally developed in previous research were used in this study (Rains et al., in press). Best practices in evaluating VPC were used in constructing the high and low VPC feedback (Samter & MacGeorge, 2016). High VPC messages explicitly acknowledged the recipient’s emotional state and attempted to reappraise their situation. Low VPC messages denied and criticized the recipient’s feelings. Although high and low VPC messages were standardized so that all participants in a given condition received the exact same feedback, the collection of individual messages met the technical criteria articulated by Samter and MacGeorge (2016) for high and low VPC. The full script for the interaction totaled 190 words and can be found in Table 1. After participants completed the interaction with ListenerBot, they were directed to a separate web page to complete a questionnaire containing measures of the dependent variables.

Table 1.
Script containing the VPC manipulation.

ListenerBot’s questions and responses to answers

Question 1 Greetings! I am ListenerBot. I am here to listen to problems. Can you please tell me about your problem--what specifically is going on?

Response 1 High VPC: That sounds like a tough situation. I’m sorry you have to deal with that. I think that anyone in your situation would find it stressful.

Low VPC: That does not sound like it would be that big of a deal for most people. It sounds like you are taking this issue way too seriously.

Question 2 How does this situation make you feel?

Response 2 High VPC: I completely understand why you would feel that way. I bet this will be a learning experience that you will benefit from in the future.

Low VPC: It seems like you are overreacting or exaggerating the problem. You seem like someone who tends to be overly dramatic about things.

Question 3 Why do you think it makes you feel that way?

Response 3 High VPC: Your feelings about this are completely justified.

Low VPC: You should just try to ignore your feelings.

Question 4 What bothers you about this situation?

Response 4 High VPC: I think that would bother just about anyone. It sounds frustrating.

Low VPC: That shouldn’t bother you. It’s really not a big deal.

Question 5 What have you done to deal with this situation?

Response 5 High VPC: After hearing about your situation, I can see why you have those feelings. I know it can be hard, but you are doing the best you can. I think that everything happens for a reason. This is just one step in the growing and learning process of life. I know that you will get through this and come out even stronger in the end.

Low VPC: After hearing about your situation, it sounds to me like you are blowing this out of proportion. There is no reason for you to be stressed about this. After all, you made those choices. It sounds like you haven’t tried hard enough to deal with this issue. Other people have stuff like this happen. You are not the only one to ever experience a problem.

Note. VPC = verbal person centeredness. The script was originally developed by Rains et al. (in press).

Measures

Emotional distress

Participants’ emotional distress was evaluated prior to and following the interaction with ListenerBot using a scale developed by Folkman and Lazarus (1985). Participants reported the degree to which their problem made them feel nine emotions: anxious, angry, sad, worried, fearful, disappointed, guilty, relieved (reverse scored), and happy (reverse scored). Ratings were made on a 7-point scale, with larger values indicating greater levels of distress. Change in emotional distress was determined by subtracting participants’ pre-interaction scores (M = 3.95, SD = 1.08, α = .81) from their post-interaction scores (M = 3.47, SD = 1.20, α = .84). Negative change scores indicated a reduction in emotional distress after the interaction with ListenerBot (M = −0.48, SD = 1.21).

Evaluations of ListenerBot

Participants’ evaluations of ListenerBot were collected after the interaction using a measure developed for this study based on previous supportive communication research (Goldsmith, McDermott, & Alexander, 2000). Participants rated the degree to which they perceived ListenerBot to be caring, supportive, comforting, helpful, and concerned. Ratings were made on a 7-point scale with larger values indicating more positive evaluations (M = 3.04, SD = 1.92, α = .96).

Discussion of their problem

The extent to which participants discussed their problem was measured using their answers to the five questions posed by ListenerBot. The Linguistic Inquiry Word Count (Pennebaker, Boyd, Jordan & Blackburn, 2015) software was used to identify the total number of words participants wrote (M = 129.46 words, SD = 72.63). Following Jones and Wirtz (2006), the proportion of positive emotion words (M = 2.80%, SD = 1.93) and negative emotion words (M = 4.40%, SD = 2.00) in participants’ responses was also examined. Positive emotion words (e.g., happy, beloved, playful) and negative emotion words (e.g., fear, anger, envious) reflect the discussion of one’s feelings or affective states. Participants’ use of insight words (e.g., understand, reason, discover) was also examined (M = 3.72%, SD = 2.13). This class of language reflects sensemaking and has been linked with beneficial outcomes of talking about stressors in several studies (e.g., Rains & Keating, 2015; Shim, Cappella, & Hahn, 2011).

Manipulation checks and control variables

Manipulation checks were administered to verify the effectiveness of the expectations prime and support quality manipulation. The check for the helpfulness prime was evaluated after the prime but before participants interacted with ListenerBot. Participants were asked to report how helpful they expected ListenerBot to be on a 7-point scale with the anchors not at all helpful and very helpful and larger values indicating greater helpfulness (M = 3.32, SD = 1.47).

The VPC manipulation check included 6 semantic differential items developed in previous research and focused specifically on the feedback participants received from ListenerBot (Jones & Guerrero, 2001). Participants rated the degree to which they perceived ListenerBot’s feedback to be inappropriate/appropriate, unsupportive/supportive, insensitive/sensitive, uncaring/caring, unhelpful/helpful, and not understanding/understanding. Ratings were made on a 7-point scale with larger values indicating that participants perceived the feedback they received to be higher quality (M = 3.86, SD = 1.86, α = .91). Finally, given sex differences in the effects of VPC documented in prior research (e.g., Burleson et al., 2005; High & Solomon, 2014), participant sex was included as a control variable. As previously reported, the sample was more likely to be female (68.8%) than male.

Results

Preliminary analyses

A series of preliminary analyses were conducted to evaluate the experimental procedures. An analysis of covariance (ANCOVA) controlling for participant sex indicated that participants who received high VPC feedback (M = 5.13, SE = .12) evaluated ListenerBot’s messages to be more effective than participants who received low VPC feedback (M = 2.59, SE = .12), F(1, 227) = 214.82, p < .001, η² = .46. A second ANCOVA controlling for participant sex revealed a significant main effect for the expectations prime on participants’ perceptions that the interaction with ListenerBot would be helpful, F(2, 227) = 25.47, p < .001, η² = .18. Post hoc pairwise comparisons showed that participants anticipated that ListenerBot would be significantly less helpful prior to the interaction in the unhelpful prime condition (M = 2.43, SE = .15) than in the no prime (M = 3.75, SE = .15) and helpful prime (M = 3.76, SE = .15) conditions. Additionally, a one-sample t-test showed that participants’ perceptions of ListenerBot’s helpfulness was below the scale midpoint in the unhelpful prime condition, t(78) = −12.95, p < .001.

However, the difference between the no prime and helpful prime conditions was not statistically significant, and a one-sample t-test showed that participants’ perceptions of ListenerBot’s helpfulness in the helpful prime condition was not significantly different from the scale midpoint, t(77) = −1.45, p = .15. Given this pattern of results, the no prime condition was excluded in testing the main effect for the helpfulness prime variable. The absence of a difference in perceived helpfulness between the helpful prime and no prime conditions undermines the value of comparing these two conditions. Additionally, readers should note that participants in the helpful prime condition did not perceive ListenerBot to be objectively helpful. Although participants were significantly more optimistic about the potential helpfulness of ListenerBot in the helpful prime condition than in the unhelpful prime condition, participants in the helpful prime condition can best be described as ambivalent about the potential of ListenerBot to actually assist them with their challenge.

Several other analyses were conducted to further evaluate the experimental procedures. ANCOVAs controlling for participant sex showed that there were no differences in participants’ initial level of emotional distress about their problem prior to interacting with ListenerBot across the three expectation prime conditions, F(1, 227) = 0.821, p = .44, η² = .007, or the two VPC conditions, F(1, 227) = 0.36, p = .550, η² = .002. Additionally, two single-item measures rated on 7-point scales were included to evaluate participants’ beliefs that ListenerBot was a computer program. The mean score for the item indicating that ListenerBot was a computer program was significantly greater than the scale midpoint (M = 6.49, SD = 0.84), t(233) = 45.56, p < .001, and the mean rating for the item indicating that ListenerBot was a human pretending to be a computer program was significantly below the scale midpoint (M = 2.14, SD = 1.37), t(233) = −20.85, p < .001. This final set of analyses offers evidence that the problem experienced by participants did not vary across conditions prior to interacting with ListenerBot and that ListenerBot was realistically perceived as a computer program.

Effects of support quality

H1 predicted that participants who received high VPC feedback would (a) perceive the support provider more positively, (b) engage in greater discussion of their problem, and (c) experience a greater reduction in emotional distress than participants who received low VPC feedback. ANCOVAs controlling for participant sex showed that participants who received high VPC feedback perceived ListenerBot more positively, F(1, 231) = 151.43, p < .001, η² = .39, and experienced a greater reduction in emotional distress, F(1, 231) = 14.00, p < .001, η² = .06, than participants who received low VPC feedback. Mean scores can be found in Table 2. H1a and H1c were supported. There were no differences between participants who received high and low VPC feedback in the total number of words used in responding to ListenerBot, F(1, 231) = 0.26, p = .611, η² = .001, use of positive emotion words, F(1, 231) = 0.07, p = .786, η² = .001, use of negative emotion words, F(1, 231) = 0.06, p = .806, η² = .001, nor use of insight words, F(1, 231) = 3.24, p = .073, η² = .01. H1b was not supported.

Table 2.
Means and standard errors for dependent measures across the support quality and expectations prime conditions.

Support quality

Low VPC High VPC

M SE M SE

Total words used 127.04 6.72 131.88 6.72

Positive emotion words 2.83 0.18 2.76 0.18

Negative emotion words 4.37 0.19 4.44 0.19

Insight words 3.47 0.20 3.97 0.20

Evaluation of ListenerBot 1.83 0.14 4.24 0.14

Change in distress −0.19 0.11 −0.77 0.11

Expectations prime

Unhelpful Helpful

M SE M SE

Total words used 113.30 7.82 137.07 7.87

Positive emotion words 2.60 0.22 2.90 0.23

Negative emotion words 4.55 0.24 4.75 0.24

Insight words 4.02 0.24 3.69 0.24

Evaluation of ListenerBot 2.36 0.21 3.45 0.21

Change in distress −0.29 0.14 −0.57 0.14

Note. VPC = verbal person centeredness. Means and standard errors are adjusted, controlling for participant sex. The values for positive emotion words, negative emotion words, and insight words are proportions relative to the total number of words used by participants.

Effects of expectations prime

H2 predicted that participants who received a prime indicating that the provider would be helpful would (a) perceive the provider more positively, (b) engage in greater discussion of their problem, and (c) experience a greater reduction in emotional distress than participants who received a prime indicating that the provider would be unhelpful. Consistent with H2a, ANCOVAs controlling for participant sex showed that participants in the helpful prime condition perceived ListenerBot more positively than participants in the unhelpful prime condition, F(1, 154) = 13.88, p < .001, η² = .08. The results related to H2b were mixed. Although participants in the helpful prime condition used a greater number of words in discussing their problem with ListenerBot, F(1, 154) = 4.57, p = .034, η² = .03, there were no statistically significant differences in their use of positive emotion words, F(1, 154) = 0.85, p = .359, η² = .005, negative emotion words, F(1, 154) = 0.35, p = .555, η² = .002, nor insight words, F(1, 154) = 0.93, p = .336, η² = .006. Finally, there was no significant difference in emotional distress reduction between the two prime conditions, F(1, 154) = 2.08, p = .152, η² = .01. H2c was not supported. The means and standard errors for these tests can be found in Table 2.

Indirect effects

H3 and H4 made predictions about indirect effects of VPC and the helpfulness prime on change in emotional distress through support seekers’ discussion of their problems and perceptions of the support provider. Contemporary procedures for testing indirect effects involving bootstrapped confidence intervals (BCIs) were followed using the PROCESS macro for SPSS (Hayes, 2013). The no prime condition was excluded from the expectations prime variable in testing the indirect effects. Sex was included as a covariate in all tests, and all reported coefficients were unstandardized.

H3a and H3b were not supported. The confidence intervals for the indirect effects of VPC on change in distress through number of words used in discussing the problem (.002, 95% BCI [−.02, .03]), use of positive emotion words (.002, 95% BCI [−.02, .03]), use of negative emotion words (−.001, 95% BCI [−.03, .02]), and use of insight words (−.002, 95% BCI [−.04, .04]) included zero. Similarly, the confidence intervals for the indirect effects of the helpfulness expectations prime on emotional distress change through number of words used in discussing the problem (.02, 95% BCI [−.05, .09]), use of positive emotion words (−.01, 95% BCI [−.06, .04]), use of negative emotion words (−.001, 95% BCI [−.04, .04]), and use of insight words (.002, 95% BCI [−.03, .05]) included zero. These patterns were inconsistent with the predictions made in H3a and H3b. VPC and the helpfulness prime did not impact participants’ distress change by influencing their efforts to discuss their problem.

H4a and H4b were supported. The confidence interval for the indirect effect of VPC on change in emotional distress through perceptions of ListenerBot did not include zero (−.63, 95% BCI [−.90, −.38]). The same trend was observed for the indirect effect of the helpfulness expectations prime on distress change through perceptions of ListenerBot (−.27, 95% BCI [−.48, −.10]). These results indicated that VPC and the helpfulness prime contributed to participants’ reduction in emotional distress by influencing their perceptions of ListenerBot. Participants who received high VPC feedback and who were in the helpful prime condition were more likely to evaluate ListenerBot positively and, in turn, benefit from the interaction.

Discussion

This project contributes to scholarship on social support in personal relationships by spotlighting the role of support seekers and examining the effects of seeker expectations and VPC during supportive interactions. The results of the experiment show that the expectations seekers have about the helpfulness of a support provider can impact supportive interactions in several ways and underscore the utility of VPC in fostering beneficial support outcomes. These findings and their implications for research on supportive communication are considered in the following paragraphs.

Expectations and support seeking

The results offer evidence that support seekers’ expectations can influence aspects of supportive interactions. Participants who received a prime indicating that ListenerBot would be helpful evaluated ListenerBot more positively and used more words in discussing their problem than participants in the unhelpful prime condition. Consistent with theorizing about behavioral confirmation (Snyder & Swann, 1978; Snyder et al., 1977), participants’ perceptions and behavior followed their expectations about the potential helpfulness of the conversation. Although the provider helpfulness prime did not directly influence the degree to which participants benefited from the interaction, it did contribute to participants’ change in distress indirectly through shaping their perceptions of ListenerBot. The results of this study contribute to scholarship on support seeking by showing that expectations can influence people’s perceptions and behavior in sharing their problems. Prior research has been limited to demonstrating that people hold expectations about the types of support they expect from trained help providers (Jay et al., 2000) and that expectations serve as the basis for evaluating received support (McDowell et al., 2010; Seiger & Wiese, 2011).

It is notable that the helpfulness prime did not directly lead to a reduction in emotional distress. Although expectations impacted participants’ perceptions and behaviors during the interaction, they did not completely override the feedback provided by ListenerBot. This finding helps to contextualize the effects of behavioral confirmation (Snyder & Swann, 1978; Snyder et al., 1977) during supportive interactions. The helpfulness of any given supportive encounter is not simply determined by those expectations a seeker brings to the interaction. Expectations play an important role in shaping interactions, but the interactions themselves—including the quality of support messages produced by providers—remain consequential. More broadly, the results of this study highlight the active role of support seekers during supportive communication. Support seekers’ expectations influenced their perceptions, volume of talk about their problem, and, indirectly, the benefits they experienced from the discussion. The findings from this study add to the growing body of research examining support seeking behavior (Barbee & Cunningham, 1995; Buehler et al., 2018; Oh & LaRose, 2016; Youngvorst & High, 2018) by highlighting the important roles played by seeker expectations and broader interactional processes in supportive communication.

Two caveats should be considered in evaluating the preceding results. First, the mean score for helpfulness expectations in the helpful prime condition did not exceed the scale midpoint. As such, the previously identified differences between prime conditions effectively tap participants’ expectations that ListenerBot would be relatively more or less unhelpful. Although it seems likely that the observed trends would extend to comparing expectations of helpfulness and unhelpfulness, our data cannot offer any definitive conclusions about this possibility. Second, the study design only allowed us to evaluate the first half of behavioral confirmation (Snyder & Swann, 1978; Snyder et al., 1977). Holding constant the supportive feedback provided in ListenerBot’s responses made it possible to isolate the effects of seekers’ expectations on their perceptions and behavior in discussing their problem. This approach, however, did not allow us to determine how a seekers’ expectations ultimately influence the responses of support providers.

VPC and support outcomes

The results of this study also replicate previous research documenting the benefits of high VPC messages (High & Dillard, 2012; Samter & MacGeroge, 2016)—including when delivered by a computer program (Rains et al., in press). Participants who received high VPC feedback evaluated ListenerBot more favorably and experienced a greater reduction in distress after their interaction than participants who received low VPC feedback. These findings contribute to research on VPC in several ways.

First, this study adds to the relatively small number of experiments that have investigated the effects of VPC during real-time interactions in which support seekers discuss their personal problems. All but two of the studies included in High and Dillard’s (2012) meta-analysis examined the effects of VPC by having participants respond to a hypothetical scenario involving what others did or how they might feel in a given situation. Although the effect of high VPC messages on distress reduction was smaller than the overall effect estimate reported by High and Dillard (2012), it was almost identical with the estimates from the two studies they identified in which the impact of high VPC messages on affective improvement was examined among people talking about their problems with a confederate support provider (r = .25). The effect of high VPC messages on distress reduction observed in this study was also similar to the results of our previous research using ListenerBot (Rains et al., in press). The consistent findings between prior studies employing confederates and this project underscore the robustness of VPC as a comforting strategy. Even when delivered in a generic format by a computer during a brief interaction, high VPC messages were more helpful than low VPC messages in reducing support seekers’ distress.

A second contribution made by this project to VPC research involves the theory of conversationally induced reappraisals (Burleson & Goldsmith, 1998) and, in particular, participants’ efforts in discussing their problems. In previous work employing confederate support providers, greater levels of VPC encouraged support seekers to use more positive emotion words, negative emotion words, and cognitive mechanism words (Jones & Wirtz, 2006). In this study, VPC did not influence participants’ use of positive emotion words, negative emotion words, insight words, nor the total number of words used in responding to ListenerBot. The absence of significant differences in emotion words is consistent with our prior research employing ListenerBot (Rains et al., in press). These findings raise questions about the importance of how support seekers talk about their problems. Despite failing to engage in greater discussion of their thoughts or feelings, participants who received high VPC feedback from ListenerBot nonetheless experienced a greater reduction in emotional distress than participants who received low VPC feedback.

It is also possible that the discrepancy between the results from prior research employing human confederates and this project could stem from the standardized nature of interactions with ListenerBot. Asking participants to respond to the same set of questions might have fostered homogeneity in responses and mitigated variation in specific types of discourse. A second possibility is that the broader connection between disclosure and emotional distress is more complex than we initially anticipated. A recursive relationship may exist whereby talk leads to a reduction in emotional distress, which leads to more talk and a further reduction in distress.

A third contribution of this study involves the role of computer programs as resources for social support. The preliminary analyses offer evidence that participants clearly perceived ListenerBot to be a computer program. Nonetheless, the benefits of high VPC feedback relative to low VPC feedback persisted. These findings highlight the tendency to employ rules for human interaction to interactions with computers (Nass & Moon, 2000; Reeves & Nass, 1996) and underscore the potential for computers to serve as a support resource (Brave et al., 2005; Liu & Sundar, 2018). Although interacting with computers is different from human support providers in important ways, the results of this study suggest there are some areas—such as the benefits of high VPC feedback—that generalize.

Limitations

The findings from this study should be considered in the light of a few limitations. One involves the nature of participants’ interaction with ListenerBot. Some readers might question the degree to which the experiment constituted a supportive interaction. Although participants’ conversations with ListenerBot were likely less extensive than in studies employing human confederates (e.g., Jones & Wirtz, 2006), they met MacGeorge et al.’s (2011) definition of supportive communication. Similarly, these conversations generally followed the four-phase structure of supportive interactions outlined by Barbee and Cunningham (1995) involving support activation, support provision, target reactions, and helper responses. Participants in the present study were engaged over multiple turns in which they were asked questions about their problem and presented with feedback to their responses.

A second and related limitation stems from the use of standardized responses containing high or low VPC feedback. Despite not being tailored to each respondent, the messages included in ListenerBot’s feedback meet the criteria for high and low VPC feedback (Samter & MacGeorge, 2016). Moreover, the VPC manipulation produced beneficial effects equivalent to what has been observed in studies employing human support providers (High & Dillard, 2012). A third limitation involves our decision to exclude moderate VPC feedback. We elected to focus on high and low VPC feedback in this study because they represent the end points of the VPC continuum. Nonetheless, moderate VPC messages are a distinct class and warrant consideration in future research (Jones, Bodie, Youngvorst, Navarro, & Danielson, 2018).

A final limitation involves the prime used to cultivate expectations about the helpfulness of ListenerBot. Expectations in the helpful prime condition did not differ from the no prime condition and, at best, only reflected an ambivalence about the potential to receive meaningful aid from ListenerBot. One explanation for this result is that participants were generally skeptical about the likelihood of a computer program to help them cope with a personal problem. The mean scores for the manipulation check measure capturing the anticipated helpfulness of ListenerBot prior to the interaction were slightly below the scale midpoint in each of the three expectation prime conditions. This may have been an artifact of the way in which ListenerBot was presented. Participants in all conditions were told that ListenerBot was developed by the researchers and still undergoing testing. It is important to note, however, that the helpful prime did generate significantly more positive expectations about the potential utility of interacting with ListenerBot than the unhelpful prime.

Conclusion

This project advances scholarship on social support in personal relationships by privileging the role of support seekers and considering the implications of seeker expectations and VPC for supportive interactions and outcomes. The results offer evidence that the expectations support seekers bring to an interaction about its helpfulness impact their subsequent perceptions and behavior. Additionally, although the benefits of VPC for coping with emotional distress were observed in human–computer interaction, the results regarding participants’ language use in discussing their problems were unexpected. Taken together, the findings from this study highlight the importance of support seeking but also emphasize the need for continued research on this topic and, more broadly, supportive communication in personal relationships.

ListenerBot’s questions and responses to answers
Question 1	Greetings! I am ListenerBot. I am here to listen to problems. Can you please tell me about your problem--what specifically is going on?
Response 1	High VPC: That sounds like a tough situation. I’m sorry you have to deal with that. I think that anyone in your situation would find it stressful.
Low VPC: That does not sound like it would be that big of a deal for most people. It sounds like you are taking this issue way too seriously.
Question 2	How does this situation make you feel?
Response 2	High VPC: I completely understand why you would feel that way. I bet this will be a learning experience that you will benefit from in the future.
Low VPC: It seems like you are overreacting or exaggerating the problem. You seem like someone who tends to be overly dramatic about things.
Question 3	Why do you think it makes you feel that way?
Response 3	High VPC: Your feelings about this are completely justified.
Low VPC: You should just try to ignore your feelings.
Question 4	What bothers you about this situation?
Response 4	High VPC: I think that would bother just about anyone. It sounds frustrating.
Low VPC: That shouldn’t bother you. It’s really not a big deal.
Question 5	What have you done to deal with this situation?
Response 5	High VPC: After hearing about your situation, I can see why you have those feelings. I know it can be hard, but you are doing the best you can. I think that everything happens for a reason. This is just one step in the growing and learning process of life. I know that you will get through this and come out even stronger in the end.
Low VPC: After hearing about your situation, it sounds to me like you are blowing this out of proportion. There is no reason for you to be stressed about this. After all, you made those choices. It sounds like you haven’t tried hard enough to deal with this issue. Other people have stuff like this happen. You are not the only one to ever experience a problem.

	Support quality
Total words used	127.04	6.72	131.88	6.72
Positive emotion words	2.83	0.18	2.76	0.18
Negative emotion words	4.37	0.19	4.44	0.19
Insight words	3.47	0.20	3.97	0.20
Evaluation of ListenerBot	1.83	0.14	4.24	0.14
Change in distress	−0.19	0.11	−0.77	0.11
	Expectations prime
Unhelpful	Helpful
	M	SE	M	SE
Total words used	113.30	7.82	137.07	7.87
Positive emotion words	2.60	0.22	2.90	0.23
Negative emotion words	4.55	0.24	4.75	0.24
Insight words	4.02	0.24	3.69	0.24
Evaluation of ListenerBot	2.36	0.21	3.45	0.21
Change in distress	−0.29	0.14	−0.57	0.14

Footnotes

Acknowledgements

The authors thank Dr Graham Bodie, who served as action editor, and the three anonymous reviewers for their helpful feedback.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Stephen A. Rains

Open research statement

As part of IARR’s encouragement of open research practices, the authors have provided the following information: This research was not pre-registered. The data used in the research can be obtained via e-mail from the first author at: srains@email.arizona.edu.

References

Barbee

A. P.

Cunningham

M. R.

(1995). An experimental approach to social support communications: Interactive coping in close relationships. Annals of the International Communication Association, 18, 381–413. doi:10.1080/23808985.1995.11678921

Bodie

G. D.

(2011). The role of thinking in the comforting process: An empirical test of a dual-process framework. Communication Research, 40, 533–558. doi:10.1177/0093650211427030

Brave

Nass

Hutchinson

(2005). Computers that care: Investigating the effects of orientation of emotion exhibited by an embodied computer agent. International Journal of Human-Computer Studies, 62, 161–178. doi:10.1016/j.ijhcs.2004.11.002

Buehler

E. M.

Peterson

A. M.

High

A. C.

(2018). Investigating the role of problem severity and channel publicness on the sophistication of support seeking messages on social network sites: A replication study. Communication Studies, 69, 304–313. doi:10.1080/10510974.2018.1463275

Burgoon

J. K.

Hale

J. L.

(1988). Nonverbal expectancy violations: Model elaboration and application to immediacy behaviors. Communication Monographs, 55, 58–79. doi:10.1080/03637758809376158

Burleson

B. R.

(1987). Cognitive complexity. In McCroskey

J. C.

Daly

J. A.

(Eds.), Personality and interpersonal communication (pp. 305–349). Newbury Park, CA: Sage.

Burleson

B. R.

(2008). What counts as effective emotional support? Explorations of situational and individual differences. In Motley

M. T.

(Ed.), Studies in applied interpersonal communication (pp. 207–227). Newbury Park, CA: Sage.

Burleson

B. R.

Goldsmith

D. J.

(1998). How the comforting process works: Alleviating emotional distress through conversationally induced reappraisals. In Andersen

P. A.

Guerrero

L. K.

(Eds.), Communication and emotion (pp. 246–275). Orlando, FL: Academic Press.

Burleson

B. R.

Holmstrom

A. J.

Gilstrap

C. M

. (2005). “Guys Can’t Say That to Guys”: Four experiments assessing the normative motivation account for deficiencies in the emotional support provided by men. Communication Monographs, 72, 468–501. doi:10.1080/03637750500322636

10.

Cacioppo

J. T.

Hawkley

L. C.

(2009). Perceived social isolation and cognition. Trends in Cognitive Sciences, 13, 447–454. doi:10.1016/j.tics.2009.06.005

11.

Cohen

Underwood

Gottlieb

(2000). Social support measurement and interventions: A guide for health and social scientists. New York, NY: Oxford.

12.

Folkman

Lazarus

R. S.

(1985). If it changes it must be a process: Study of emotion and coping during three stages of a college examination. Journal of Personality and Social Psychology, 48, 150–170. doi:10.1037/0022-3514.48.1.150

13.

Goldsmith

D. J.

McDermott

V. M.

Alexander

S. C.

(2000). Helpful, supportive and sensitive: Measuring the evaluation of enacted social support in personal relationships. Journal of Social and Personal Relationships, 17, 369–391. doi:10.1177/0265407500173004

14.

Goldsmith

D. J.

(2004). Communicating social support. New York, NY: Cambridge University Press.

15.

Hayes

A. F

. (2013). Introduction to mediation, moderation, and conditional process analysis. New York, NY: Guilford Press.

16.

High

A. C.

Dillard

J. P.

(2012). A review and meta-analysis of person-centered messages and social support outcomes. Communication Studies, 63, 99–118. doi:10.1080/10510974.2011.598208

17.

High

A. C.

Scharp

K. M.

(2015). Examining family communication patterns and seeking social support direct and indirect effects through ability and motivation. Human Communication Research, 41, 459–479. doi:10.1111/hcre.12061

18.

High

A. C.

Solomon

D. H.

(2014). Communication channel, sex, and the immediate and longitudinal outcomes of verbal person-centered support. Communication Monographs, 81, 439–468. doi:10.1080/03637751.2014.933245

19.

Holmstrom

A. J.

Clare

D. D.

Russell

(2014). Problem-focused content in the job search: Two tests of the cognitive-emotional theory of esteem support messages. Human Communication Research, 40, 161–187. doi:10.1111/hcre.12025

20.

Jay

L. R.

Afifi

W. A.

Samter

(2000). The role of expectations in effective genetic counseling. Journal of Genetic Counseling, 9, 95–116. doi:10.1023/A:1009493424814

21.

Jones

S. M.

Guerrero

L. K.

(2001). The effects of nonverbal immediacy and verbal person centeredness in the emotional support process. Human Communication Research, 27, 567–596. doi:10.1111/j.1468-2958.2001.tb00793.x

22.

Jones

S. M.

Bodie

G. D.

Youngvorst

Navarro

Danielson

(2018). Mapping the terrain of person-centered supportive conversations. Communication Monographs, 85, 467–490, doi:10.1080/03637751.2018.1501503

23.

Jones

S. M.

Wirtz

J. G.

(2006). How does the comforting process work? An empirical test of an appraisal-based model of comforting. Human Communication Research, 32, 217–243. doi:10.1111/j.1468-2958.2006.00274.x

24.

Lazarus

R. S.

(1999). Stress and emotion: A new synthesis. New York, NY: Springer.

25.

Lazarus

R. S.

Folkman

(1984). Stress, appraisal, and coping. New York, NY: Springer.

26.

Liu

Sundar

(2018). Should machines express sympathy and empathy? Experiments with a health chatbot. Cyberpsychology, Behavior, and Social Networking, 21, 625–636. doi:10.1089/cyber.2018.0110

27.

MacGeorge

E. L.

Feng

Burleson

(2011). Supportive communication. In Knapp

M. L.

Daly

J. A.

(Eds.), Handbook of interpersonal communication (pp. 317–354). Thousand Oaks, CA: Sage.

28.

MacGeorge

E. L.

Feng

Guntzviller

L. M.

(2016). Advice: Expanding the communication paradigm. Annals of the International Communication Association, 40, 213–243. doi:10.1080/23808985.2015.11735261

29.

McDowell

M. E.

Occhipinti

Ferguson

Dunn

Chambers

S. K.

(2010). Predictors of change in unmet supportive care needs in cancer. Psycho-Oncology, 19, 508–516. doi:10.1002/pon.1604

30.

Merton

R. K.

(1948). The self-fulfilling prophecy. The Antioch Review, 8, 193–210. doi:10.2307/4609267

31.

Nass

Moon

(2000). Machines and mindlessness: Social responses to computers. Journal of Social Issues, 56, 81–103. doi:10.1111/0022-4537.00153

32.

H. J.

LaRose

(2016). Impression management concerns and support-seeking behavior on social network sites. Computers in Human Behavior, 57, 38–47. doi:10.1016/j.chb.2015.12.005

33.

Pennebaker

J. W.

Boyd

R. L.

Jordan

Blackburn

(2015). The development and psychometric properties of LIWC2015. Austin, TX: University of Texas at Austin.

34.

Rains

S. A.

Keating

D. M.

(2015). Health blogging: An examination of the outcomes associated with making public, written disclosures about health. Communication Research, 42, 107–133. doi:10.1177/0093650212458952

35.

Rains

S. A.

Pavlich

C. A.

Tsetsi

Akers

. (in press). Evaluating the scope and theoretical mechanisms of person-centered social support: The case of the comforting computer program. Communication Research [advanced online publication]. doi:10.1177/0093650219852554

36.

Reeves

Nass

(1996). The media equation: How people treat computers, television, and media like real people and places. Cambridge, UK: Cambridge University Press.

37.

Samter

MacGeorge

E. L.

(2016). Coding comforting behavior for verbal person centeredness. In Canary

D. J.

VanLear

(Eds.), Researching communication interaction behavior: A sourcebook of methods and measures (pp. 107–128). Thousand Oaks, CA: Sage.

38.

Seiger

C. P.

Wiese

B. S.

(2011). Social support, unfulfilled expectations, and affective well-being on return to employment. Journal of Marriage and Family, 73, 446–458. doi:10.1111/j.1741-3737.2010.00817.x

39.

Shim

Cappella

J. N.

Han

J. Y.

(2011). How does insightful and emotional disclosure bring potential health benefits? Study based on online support groups for women with breast cancer. Journal of Communication, 61, 432–454. doi:10.1111/j.1460-2466.2011.01555.x

40.

Snyder

Swann

W. B.

(1978). Hypothesis-testing processes in social interaction. Journal of Personality and Social Psychology, 36, 1202–1212. doi:10.1037/0022-3514.36.11.1202

41.

Snyder

Tanke

E. D.

Berscheid

(1977). Social perception and interpersonal behavior: On the self-fulfilling nature of social stereotypes. Journal of Personality and Social Psychology, 35, 656–666. doi:10.1037/0022-3514.35.9.656

42.

Thoits

P. A.

(2011). Mechanisms linking social ties and support to physical and mental health. Journal of Health and Social Behavior, 52, 145–161. doi:10.1177/0022146510395592

43.

Tong

S. T.

Walther

J. B.

(2015). The confirmation and disconfirmation of expectancies in computer-mediated communication. Communication Research, 42, 186–212. doi:10.1177/0093650212466257

44.

Youngvorst

L. J.

High

A. C.

(2018). “Anyone free to chat?” Using technology features to elicit quality support online. Communication Monographs, 85, 206–223. doi:10.1080/03637751.2018.1426871

	Support quality
	Low VPC		High VPC
	M	SE	M	SE
Total words used	127.04	6.72	131.88	6.72
Positive emotion words	2.83	0.18	2.76	0.18
Negative emotion words	4.37	0.19	4.44	0.19
Insight words	3.47	0.20	3.97	0.20
Evaluation of ListenerBot	1.83	0.14	4.24	0.14
Change in distress	−0.19	0.11	−0.77	0.11
	Expectations prime
	Unhelpful		Helpful
	M	SE	M	SE
Total words used	113.30	7.82	137.07	7.87
Positive emotion words	2.60	0.22	2.90	0.23
Negative emotion words	4.55	0.24	4.75	0.24
Insight words	4.02	0.24	3.69	0.24
Evaluation of ListenerBot	2.36	0.21	3.45	0.21
Change in distress	−0.29	0.14	−0.57	0.14