Electronic Thesis and Dissertation Repository

Thesis Format

Integrated Article


Doctor of Philosophy




Heerey, Erin


We often initiate social relationships with others through revelations of personal information, or self-disclosure. Self-disclosure is heavily involved in shaping interpersonal liking, but there are disparate and sometimes contradictory findings in the literature regarding the causal relationship between them. Moreover, a lack of careful control in experimental designs in many existing studies failed to eliminate important confounding factors that might provide alternative explanations for the disclosure-liking relationship. Here, we examined the relationships between self-disclosure and interpersonal liking during initial social interactions, while carefully controlling for a potential confounding factor, similarity between the social partners.

Across the first five experiments, I independently manipulated disclosers’ self-disclosure depth, i.e., how personal and intimate the disclosures are, and their self-disclosed similarity with their social partners. High self-disclosed similarity was consistently found to lead to greater initial liking of a discloser. In comparison, the experiments failed to find support for the idea that people favor those who self-disclose more deeply, as suggested in the literature. In Experiment 6, I manipulated initial liking within a set of social partners and successfully replicated another disclosure-liking relationship identified in the literature, namely, the effect that people self-disclose to a greater extent to those whom they like. It was also found that, contrary to the expectation, participants’ risk-taking tendencies negatively predicted their self-disclosure depth to others. In Experiment 7, I extended the investigation to an emerging and novel social context and examined how self-disclosed similarity from an Artificially Intelligent (AI) agent influenced people’s perceptions of and responses to the agent. A significant interaction between the perceived identity of the partner (i.e., AI versus human) and level of self-disclosed similarity was found. The results were interpreted in light of the “uncanny valley effect”, which suggests that a high level of human realism displayed by an automatic agent could elicit unpleasant or “eerie” feelings.

Through this series of experiments, I iteratively developed the paradigm to more closely mimic real-world social disclosures. The findings help disentangle the causal relationship between self-disclosure and initial liking and provide insights into some of the subtleties and processes underlying relationship formation.

Summary for Lay Audience

Making friends is important. Being able to enjoy good social relationships with other people is beneficial to both our psychological and physical health. The friend-making process frequently starts when we tell each other information about ourselves, such as our past, hobbies, thoughts, and feelings. This act of revealing our own information to another person is called self-disclosure.

What should we self-disclose in our first interaction with another person to best kindle the budding friendship? The past literature suggests that self-disclosing deeper, rather than more superficial, information about yourself might make the other person like you more. However, there are some methodological problems with the previous studies that render this conclusion questionable. Specifically, the fact that you self-disclose more deeply to the other person and that this person likes you more might both have resulted from a greater similarity between the two of you.

To address this issue, we investigated whether self-disclosing more deeply to a stranger makes them like oneself more, after experimentally controlling for the level of similarity between the two people. We consistently found people to like those who self-disclosed a greater similarity to themselves and not those who self-disclosed more deeply to them. In other words, revealing to a stranger that you are similar to them would make them like you more, whereas telling them deeply personal information about yourself would likely not. I also investigated whether the causal relationship is the other way around, namely, whether liking the other person more to start with leads one to self-disclose more deeply to them. My findings supported this account. Finally, we explored whether people like an artificially intelligent (AI) agent more if the agent self-disclosed greater similarity with themselves. Findings suggested that people reacted differently to self-disclosed similarity coming from an AI partner versus a human partner.