Classifying values in informal communication: Adapting the meta-inventory of human values for tweets



This study builds upon and expands the Meta-Inventory of Human Values (MIHV) proposed by Cheng and Fleischmann (2010) for application in informal communication settings such as Twitter. A corpus of 5,313 tweets from 32 individuals during a three-week period in March/April 2011 was analyzed using thematic analysis. In addition to the 16 value categories in the MIHV, two new categories emerged within the informal communication context – connectedness and comfort.


Values serve as “a link between self and society” (Rokeach, 1973) and are understood to be an important precursor to behaviors and decision making at both the individual and societal levels (Schwartz, 2007). Survey methodology and content analysis are two common approaches for studying human values (Fleischmann et al., 2009). This study identifies human values expressed through online content similar to other attempts to identify values through user- generated content online (Morgan, Mason, & Nahon, 2011). Specifically, this study focuses on Twitter users and their tweets.

Research on human values has received renewed attention in the 21st century from scholars in the social sciences with regard to cultural values, gender, socioeconomic status, and identity (e.g., Hitlin, 2003; Schwartz, 2007). With particular regard to the continued shift from industrialization to informationalization of Western culture and society, information scientists have taken up the task of identifying values embedded in digital content (e.g., Cheng et al., 2010; Fleischmann et al., 2009; Fleischmann, Templeton, & Boyd-Graber, 2011; Ishita et al., 2010; Morgan et al., 2011; Templeton, Fleischmann, & Boyd-Graber, 2011).

The Meta-Inventory of Human Values (MIHV) synthesizes definitions across the values-based research literature and provides a holistic and operational definition of values “as guiding principles of what people consider important in life and how something ought to be” (Cheng & Fleischmann, 2010). The MIHV was built using 12 leading value inventories from the fields of psychology, sociology, anthropology, management, advertising, and human- computer interaction. The MIHV includes only values that were listed within at least 5 of the 12 inventories. Cheng and Fleischmann have continued to evolve the MIHV, modifying terminology (e.g., accomplishment is now achievement, self-respect is now identity, and creativity is now innovation) and category descriptions to improve inter- coder agreement (Cheng & Fleischmann, Under Review) and they have explored computational methods for content analysis building from this work (Fleischmann et al., 2009; Ishita et al., 2010).

Prior to this study, the MIHV was tested and modified based on coding of prepared testimonies from the Net neutrality debate presented in front of Congressional committees and the Federal Communications Commission (FCC) (Cheng et al., 2010). The Net neutrality corpus investigated in that study consists of formal communication that is prepared in advance for a known and highly educated audience. To ensure that the MIHV is relevant for other types of communication, it is important to investigate which additional values occur in informal communication contexts like online social media sites such as Twitter.

This paper reports the findings of a thematic analysis (Braun & Clarke, 2006) of human values expressed in 1,717 of 5,313 public tweets from 32 individuals based in the United States. The study sought to address the following research questions:

  • 1What human values are expressed in informal communication, such as tweets?
  • 2How do these values differ from values expressed in formal communication?


Tweets are typically undirected and informal units of public (typically) online communication. They are not facilitated by a researcher in an interview or focus group process, nor are they dictated by official guidelines or rhetorical parameters other than a character limit.

The content of tweets, like other forms of informal communication, is affected by the environmental context of the individuals generating the content. These contexts include social (e.g., news media, popular culture, current events) and physical (e.g., weather, health) issues. User- generated content on Twitter especially is known to follow major societal trends (Kwak et al., 2010). Due to the point- in-time nature of this study, it is important to reflect on some of the major events that occurred during the March 25-April 16, 2011, time period, which may have an impact on tweets and the extent to which they are value-laden:

  • College Basketball's March Madness

  • Charlie Sheen's falling out with NBC

  • Potential government shutdown over the budget

  • Major League Baseball's Opening Day

  • Aftermath of Japanese earthquake and tsunami

  • April Fool's Day

  • Tax preparation season


Thematic analysis was used in this study because it is non- invasive; it affords access to a population that may be hard- to-reach offline, but can be found online; does not involve self-selection bias; mitigates issues of social desirability that often occurs in interview and survey techniques; and is useful for studying values, beliefs, and attitudes (Braun & Clarke, 2006; Cheng et al., 2010; Fleischmann et al., 2009).

Data Sampling

User-generated content from Twitter (, a microblogging website, was chosen for this study. Twitter displays large quantities of public communication among a wide variety of users, and the user-generated content comes in the form of an economical 140 characters of text, which is helpful when manually coding a large dataset.

Data Collection and Management

Google Reader, a web-based aggregator, was used to collect and read the Really Simple Syndication (RSS) feeds of the tweets from 50 Twitter users for a three-week period starting March 25, 2011, and ending April 16, 2011. This analysis was part of a larger study comparing how human values are expressed by Twitter users who self-identify as homeless to those who do not. One goal of that study is to better understand the needs and values of this technologically underserved audience and consider how they might impact the design of social media platforms like Twitter or new sites that might be developed in the future. This larger study was designed to use the MIHV for analysis. However, as a first step, it was first necessary to ensure that the MIHV was relevant for this population and platform. As such, thematic analysis was first conducted to identify any human values expressed via social media that could be viewed as missing from the MIHV.

Since the eventual goal of the larger project was to use the MIHV for content analysis of tweets, it was first important to identify which tweets could be viewed as codable. For the purposes of the study, only original tweets written by human users were deemed codable for content analysis. This criterion excludes re-tweets (reposting another user's tweet, symbolized by an “RT” in the message content), because re-tweets typically involve forwarding the communication and thus potentially the values of another individual. Similarly, at-replies or at-mentions (the use of the “@” sign plus another user's Twitter username at the beginning of a tweet, typically used to reply to another tweet) at the beginning of a tweet signal one-to-one rather than one-to-many or broadcast communication. Such communication may reflect not only the values of the writer but also the values of the audience (Templeton & Fleischmann, 2011). Announcement tweets from third-party applications such as Foursquare (a location-based social network), Pandora (an Internet radio application), and ThatCanBeMyNextTweet (an application that auto generates new tweets by combining parts of an individual's recent tweets) were also removed from the corpus because they were not generated manually by human users. And, finally, redundant tweets (reposted multiple times by the same user) were removed.

A total of 18 Twitter users out of the original 50 did not produce any original tweets during the three-week data collection period, resulting in a total of 5,313 tweets, with 1,717 of the tweets meeting the criteria for codable tweets, from a total of 32 Twitter users.

Data Analysis

The coding handbook for the MIHV outlines specific rules and guidelines for identifying statements that are value- laden; for coding value judgments in negative statements; and for identifying the value object and the underlying values in order to identify whether or not each value was expressed in a coding unit (Cheng & Fleischmann, Under Review). The thematic analysis used here followed the rules outlined for the MIHV, but allowed for the inductive development of new themes that represented values not classified within the MIHV that might be more likely to occur in informal communication such as tweets.


The first research question asks: What human values are expressed in informal communication, such as tweets? Values were expressed in tweets through opinions, judgments, and sarcastic statements. It was not expected that every tweet would contain an expressed value, and it was possible for some tweets to contain more than one expressed value.

For tweets or parts of tweets that did not express a value or that could not otherwise be categorized using the coding rubric, four information behavior-related categories emerged, including information sharing, information seeking, self-promotion, and self-disclosure. These categories shared similarities with groupings developed by Java et al. (2007) in an early study of Twitter use motivations and user types, though the categories described here are behavior focused. For instance, the category of information sharing combines their two categories of sharing information/URLs and reporting news, and the behavioral category of information seeking is evoked by their user type of information seeker. Daily chatter, or instances of talking about what a person is currently doing (e.g., at the grocery store, in the emergency room) was coded as self-disclosure. Finally, self-promotion, or using Twitter to link to or announce things like new blog posts, talks, or upcoming lectures emerged (Golbeck, Grimes, & Rogers, 2010).

The second research question asks: How do these values differ from values expressed in formal communication? Through the inductive coding process, two new themes, or values, were added to the MIHV: connectedness and comfort.

Connectedness was defined as: “A desire or concern with connecting to other individuals; evoking familiarity with a group of other individuals by broadcasting greetings; concerned with familial ties or friendships” (see examples in Discussion below). Connectedness is a natural value category to find in the context of social media, given that building social networks (i.e. connections) of social capital is one of the major functions of social media (e.g., Burke, Kraut & Marlow, 2011; Ellison, Steinfield, & Lampe, 2007).

Comfort was defined as: “Concerned with comfort and leisure activities; impact of environmental factors such as the weather on an individual's comfort” (see examples in Discussion below). Comfort also seems natural in an informal communication context like Twitter given the frequency of self-disclosure that appears in tweets. Often these are focused on comfort issues, such as a discussion of the weather or going on vacation. Thus, these two value categories can be added to the MIHV when studying informal communication from social media sites.


A wide range of value expressions emerged in this study. Two value categories emerged specific to the informal communication context that should be considered when analyzing similar media content in future studies. Although greetings and introductory statements were considered perfunctory in the analysis of values in the Net neutrality debate, in Twitter they were a behavior that expressed the value of social connectedness. Comments like “What's new my twits?,” “Thanks to my new followers,” and “Morning!,” when expressed through a publicly broadcast tweet show a consideration of connectedness. This value category shares similarities to sense of belonging in the Schwartz Value Survey (1994), the List of Values inventory by (Kahle, Poulos, & Sukhdial, 1988), and belonging in the Life Values Inventory by (Crace & Brown, 2002). In this study it appeared as the second highest most-expressed value (see Table 2).

Tweets expressing the value of comfort suggested the use of Twitter as a communication device for expressing the need for, or appreciation of, creature comforts and leisure. Tweets expressed this value through statements related to enjoying the weather (“Happy Saturday, how is everyone tonight? Hope you are well and the weather is nice. #luv #homeless”), having or desiring a comfortable place to sleep (“Who wants to share with others just how hard it is to sleep in their vehicle….ready go”), and through a need for leisure time (“Back in Seattle and so excited for a few days of vacation with the family we need it bad”). This value category shares similarities with leisure in the Personal Value Questionnaire (England, 1967), a comfortable life in the Rokeach Value Survey (Rokeach, 1973), and enjoying life in the Schwartz Value Survey (Schwartz, 1994). Although it was not one of the most expressed values in this study, comfort appears more often than responsibility, social order, and freedom (see Table 2).

This study suggests that researchers who use the MIHV should be aware that the media or context of the content under consideration may have an effect on the types of values expressed. The importance of context and its impact on the “local expressions of values” by an individual or group is further discussed by Le Dantec, Poole, and Wyche (2009). In their paper they presented three empirical case studies which highlighted the importance of not only context, but also the need to not over-privilege a pre- defined set of values. Our study supports this argument and thus, a combination of inductive and deductive coding may be most effective for eliciting human values across communication contexts.


This study explored the expression of human values through Twitter, one of many growing social media sites that support informal communication. The findings suggest the addition of two value categories, connectedness and comfort, to Cheng and Fleischmann's (2010) Meta- Inventory of Human Values for analysis of such content.


We would like to thank An-Shou Cheng for sharing materials about the MIHV, Jeffrey DiScala and Lloyd Beers for constructive feedback on earlier drafts of this paper, and Joseph Maki for his assistance in data management.