Many conversational agents (CAs) are developed to answer users' questions in a specialized domain. In everyday use of CAs, user experience may extend beyond satisfying information needs to the enjoyment of conversations with CAs, some of which represent playful interactions. By studying a field deployment of a Human Resource chatbot, we report on users' interest areas in conversational interactions to inform the development of CAs. Through the lens of statistical modeling, we also highlight rich signals in conversational interactions for inferring user satisfaction with the instrumental usage and playful interactions with the agent. These signals can be utilized to develop agents that adapt functionality and interaction styles. By contrasting these signals, we shed light on the varying functions of conversational interactions. We discuss design implications for CAs, and directions for developing adaptive agents based on users' conversational behaviors.

Employee engagement is becoming a critical component of an organization's success. For many companies, increases in engagement mean increases in productivity. While organizations have long used traditional questionnaires to measure engagement, these leave room for interpretation; and trying to get clarity makes the questionnaires longer, which decreases participation. On the other hand, user-generated content can often act as a barometer on employee engagement, but does not lend itself easily to engagement prediction. We present a model leveraging machine-learning techniques to analyze and predict employee engagement, using both user-generated social analytics and text content from organization-wide surveys, to calculate a real-time engagement level for employees. Our method helped identify specific words and phrases used by individual employees that play an important role in engagement prediction. This study was able to uncover critical themes driving employee engagement and behavior at the enterprise level. Our model for textual analysis cannot only predict engagement, but also provide highly valuable, and actionable, data for business leaders about the health of their organizations.

Enterprise social networking (ESN) platforms have been implemented by organizations to support employees' collaboration. However, there has been a reported lack of user participation on these platforms. This study focuses on interventions that can be used to overcome barriers of usage. A case study was conducted at IBM to investigate the effect of several interventions to positively influence ESN acceptance within IBM's Human Resources (HR) function. In this study we combined two rounds of interview data with activity log data. We interviewed the initiators of interventions, and we interviewed the employees targeted by those interventions. The activity logs allowed us to contrast actual usage of the ESN by employees from HR vs. other organizational functions. Compared to other ESN studies this study explicitly identifies factors driving or inhibiting ESN use, and shows how these factors can be addressed by interventions to significantly increase ESN use. Our results can inform and enable other organizations interested in the design of interventions to increase the use of ESN.

Behavioral statescan be transferred to others, leading people to behave in ways similar to those around them. Can this phenomenon of behavioral contagion be seen in the workplace? Using employees' organizational social media data and their workplace hierarchical network structure, we studied contagion across a large multinational corporation, focusing on an important workplace behavior - employee engagement. We measured employees' engagement based on their word choice in organizational social media, and we applied a longitudinal statistical technique which controls for homophily, employees' traits and prior expressions of engagement. We found that engagement and disengagement spread from one employee to another with direct peers exerting the strongest influence. While engagement-spread was more powerful laterally among people at the same organizational level, disengagement-spread followed the vertical managerial chain. Further, we found that disengaged co-workers exerted a stronger influence on employee's future engagement compared to the engaged co-workers. Our results suggest the need for organizations to sense and address workplace disengagement promptly. Moreover, our findings offer opportunities for using workplace interventions to promote engagement and mitigate disengagement.

Personal agent software is now in daily use in personal devices and in some organizational settings. While many advocate an agent sociality design paradigm that incorporates human-like features and social dialogues, it is unclear whether this is a good match for professionals who seek productivity instead of leisurely use. We conducted a 17-day field study of a prototype of a personal AI agent that helps employees find work-related information. Using log data, surveys, and interviews, we found individual differences in the preference for humanized social interactions (social-agent orientation), which led to different user needs and requirements for agent design. We also explored the effect of agent proactive interactions and found that they carried the risk of interruption, especially for users who were generally averse to interruptions at work. Further, we found that user differences in social-agent orientation and aversion to agent proactive interactions can be inferred from behavioral signals. Our results inform research into social agent design, proactive agent interaction, and personalization of AI agents.

Grounded Theory Method (GTM) and Machine Learning (ML) are often considered to be quite different. In this note, we explore unexpected convergences between these methods. We propose new research directions that can further clarify the relationships between these methods, and that can use those relationships to strengthen our ability to describe our phenomena and develop stronger hybrid theories.

Employee engagement is a reflection of an employee's experience of work. Previous research has analyzed each employee's experience in terms of individual factors. We provide the first report of the influence on engagement of peers (who report to the same manager) and friends (who share social ties in an internal social network), using linear regression to model employee engagement in a sample of more than 44,000 employees. We show that an employee's engagement is associated with the engagement of her/his peers, friends, and manager. Our results contribute to analyses of social factors at work, and argue for revisions to existing theories of employee engagement.

Employees increasingly are expressing ideas and feelings through enterprise social media. Recent work in CHI and CSCW has applied linguistic analysis towards understanding employee experiences. In this paper, we apply dictionary based linguistic analysis to measure 'Employee Engagement'. Employee engagement is a measure of employee willingness to apply discretionary effort towards organizational goals, and plays an important role in organizational outcomes such as financial or operational results. Organizations typically use surveys to measure engagement. This paper describes an approach to model employee engagement based on word choice in social media. This method can potentially complement surveys, thus providing more real-time insights into engagement and allowing organizations to address engagement issues faster. Our results predicting engagement scores on a survey by combining demographics with social media text demonstrate that social media text has significant predictive power compared to demographic data alone. We also find that engagement may be a state than a stable trait since social media posts closer to the administration of the survey had the most predictive power. We further identify the minimum number of social media posts required per employee for the best prediction.

The use of social media at work is gaining traction, and there is evidence to suggest that various benefits accrue from its use. Yet the relationship between using social media at work and employee performance is not clear. Through a study of 75,747 employees of a large global company over the course of 3 years, we find that some social media usage (number of forum posts, forum post length, and status update length) was positively associated with performance ratings. This study is one of the first to show the relationship among different forms of social media use and employee performance ratings.

The rise of social media in the enterprise has enabled new ways for employees to speak up and communicate openly with colleagues. This rich textual data can potentially be mined to better understand the opinions and sentiment of employees for the benefit of the organization. In this paper, we introduce Enterprise Social Pulse (ESP) -- a tool designed to support analysts whose job involves understanding employee chatter. ESP aggregates and analyzes data from internal and external social media sources while respecting employee privacy. It surfaces the data through a user interface that supports organic results and keyword search, data segmentation and filtering, and several analytics and visualization features. An evaluation of ESP was conducted with 19 Human Resources professionals. Results from a survey and interviews with participants revealed the value and willingness to use ESP, but also surfaced challenges around deploying an employee social media listening solution in an organization.

This paper presents the first in-depth evaluation of a large multi-format virtual conference. The conference took place in an avatar-based 3D virtual world with spatialized audio, and had keynote, poster and social sessions. We studied it by drawing on logs, a survey and interviews with 30 participants. We develop a model - Coalescence, Focused Interaction, Remixing (CoFIRe) -- of large synchronous interactions, and use it to discuss how the technology supported, or failed to support, the interactions that are the raison d'etre of conferences. We conclude by discussing the prospects for such large virtual gatherings.

While there are tools that allow distributed teams to collaboratively edit in real time, little work examines this practice among real teams doing real work. Even less is known about how teams from different countries make use of real-time collaborative editing tools. The current work highlights results from a qualitative user study of real-world Japanese and U.S. distributed work teams who used LiveDeck, a real-time slide editing and whiteboarding tool. Through the implementation of various novel features used as probes, differences in behavior and attitudes between team members were uncovered. Differences in the use of slide navigation options, anonymity features, and pop-up 'emotes' representing nonverbal gestures are discussed.

There is a growth in the popularity of social file sharing systems. This paper describes the design of Cattail, a social file sharing system for the enterprise. Through a 'Recent Events' stream, Cattail supports social navigation and exploratory search by inferring relevant social connections rather than purely relying on user-specified contacts. Social navigation is further supported through pivot browsing from a consolidated history of user actions on an individual's files. Through usage log analysis over an 8-month period, we found that Cattail's novel network inference and social navigation features enabled a net gain of clickthroughs. Interviews with users revealed that this led to increased discovery of relevant people and content. We conclude with a discussion of several possible enhancements to the system. Findings from our research provide a strong foundation for the design of social file sharing systems for enterprise settings.

Most knowledge repositories focus on the role of knowledge-creators. In this paper, by contrast, we examined the work of Lurkers in an enterprise file-sharing service, and we compared their lurking behaviors to the lurking behaviors of users who uploaded files (Uploaders), and users who contributed metadata about files (Contributors). For comparability, we restricted our analyses to the consuming behaviors that are common to the three roles (Uploaders, Contributors, and Lurkers). Independent principal components analysis showed highly similar seven-factor solutions of lurking activities across all three roles, although the relative emphases of those factors varied across roles. Uploaders tended to view and download more groups of files, showed less emphasis on searching for files, and tended to work directly with the file-sharing application, unmediated by external applications. Contributors showed the opposite pattern: more emphasis on searching and responding to recommendations from other users, often via a form of remote access. Lurkers' lurking behaviors were less intense, and showed little difference in emphases among the lurker factors. We use these results, and the published research literature, to motivate a research agenda for lurkers in social media.

The difficulties remote participants of distributed meetings face are widely recognized. In this paper we describe the design of an avatar-based e-meeting support tool named Olympus, which aims to ameliorate some of the challenges remote participants face in distributed meetings. Olympus provides a customizable peripheral display on the bottom of existing e-meeting solutions. An initial observational study was conducted of the use of Olympus in 6 meetings, three each of a status meeting and a presentation meeting. By illustrating how avatars were used in the two meeting types, we hope to surface design issues and refine our understanding of how avatars may be useful in the design of online meeting spaces.

Contemporary work increasingly involves interacting with strangers in technology-mediated environments. In this context, we come to rely on digital artifacts to infer characteristics of other people. This paper reports the results of a study conducted in a global company that used expertise search as a vehicle for exploring how people interpret a range of information available in online profiles in evaluating whom to interact with for expertise. Using signaling theory as a conceptual framework, we describe how certain 'signals' in various social software are hard to fake, and are thus more reliable indicators of expertise. Multi-level regression analysis revealed that participation in social software, social connection information, and self-described expertise in the corporate directory were significantly helpful in the decision to contact someone for expertise. Qualitative analysis provided further insights regarding the interpretations people form of others' expertise from digital artifacts. We conclude with suggestions on differentiating various types of information available within online profiles and implications for the design of expertise locator/recommender systems.

It is well established that there is a need to find experts to get answers or advice. A variety of expertise locator tools have emerged to help locate the right person. But there is little systematic study on what people are really looking for when such systems are used and how external factors such as job role may shape that search. We conducted a study of 75 employees who were current users of an expertise locator system. An analysis of the reasons for their search revealed that people in client facing roles are primarily seeking to have a dialog with an expert, while others are just as likely to seek answers to technical questions. We also surveyed various tools for finding experts and found that corporate directories and personal networks were most often cited as alternatives to an expertise locator. We discuss the implications of these results for the design of tools for finding experts and expert knowledge.

Expertise locator systems have been designed to help find experts within organizations. While there are many examples of these systems in the literature, there has not been any systematic analysis of the factors that predict whether a particular expertise search result will be selected for further exploration. This paper describes a study of 67 employees from 21 countries that performed a specific expertise search to find an expert using an expertise locator system. Rank order and social connection information displayed in snippets of search results were found to significantly predict whether a user considers a particular search result for further exploration. Implications for the design of expertise location systems and future research directions are discussed.

We describe the design and evaluation of K-net, a social matching system to help people learn 'who knows what' in an organization by matching people with skills with those who need them. Transactive memory theory predicts that K-net will improve individuals' awareness of 'who knows what'. This should lead to improved performance through sharing knowledge across group boundaries. We evaluate K-net in terms of these predictions in an experiment with 41 students in seven groups working on software engineering projects. Accurate recommendations improved awareness of 'who knows what' versus 'random' recommendations, but did not improve performance. Our results highlight issues related to the evaluation of systems for sharing knowledge across group boundaries.

Under what circumstances might a group member be better off as a long-distance participant rather than collocated? We ran a set of experiments to study how partially-distributed groups collaborate when skill sets are unequally distributed. Partially distributed groups are those where some collaborators work together in the same space (collocated) and some work remotely using computer-mediated communications. Previous experiments had shown that these groups tend to form semi-autonomous 'in-groups'. In this set of experiments the configuration was changed so that some player skills were located only in the collocated space, and some were located only remotely, creating local surplus of some skills and local scarcity of others in the collocated room. Players whose skills were locally in surplus performed significantly worse. They experienced 'collocation blindness' and failed to pay enough attention to collaborators outside of the room. In contrast, the remote players whose skills were scarce inside the collocated room did particularly well because they charged a high price for their skills.

Modern workplaces often bring together virtual teams where some members are collocated, and some participate remotely. We are using a simulation game to study collaborations of 10-person groups, with five collocated members and five isolates (simulated 'telecommuters'). Individual players in this game buy and sell 'shapes' from each other in order to form strings of shapes, where strings represent joint projects, and each individual players' shapes represent their unique skills. We found that the collocated people formed an in-group, excluding the isolates. But, surprisingly, the isolates also formed an in-group, mainly because the collocated people ignored them and they responded to each other.