Emotional Labor of Software Engineers Alexander Serebrenik Eindhoven University of Technology, The Netherlands, a.serebrenik@tue.nl Abstract—The concept of emotional labor, introduced by conduct on GitHub [7], the major platform open-source soft- Hochschild in 1983, refers to the “process by which workers ware development. Examples of positive behavior encouraged are expected to manage their feelings in accordance with orga- by the Contributor Covenant include “gracefully accepting nizationally defined rules and guidelines”. For instance, judges are expected to appear impartial, nurses—compassionate and constructive criticism” and “showing empathy towards other police officers—authoritative. While software development has community members”, i.e., to suppress negative emotions that been traditionally stereotyped as a nerdy “lone wolf” job less might have been triggered by criticism and amplify positive likely to induce emotional labor, nowadays software developers emotions towards the colleagues. Codes of conduct in open- become more and more social, on the one hand, and are subject source projects are experienced as problematic by certain to increasing amount of behavioral expectations, e.g., formulated as codes of conduct. software developers as witnessed by the opposing efforts In this position paper we stress that software developers are known as “No Code of Conduct”2 . Finally, exhaustion related subject to emotional labor, envision how emotional labor can be to emotional labor has been shown to be one of the most identified based on emotion detection techniques applied in soft- important variables explaining IT career abandonment [8]. ware engineering, suggest possible antecedents and consequents While attention to emotions expressed by developers is of emotional labor and discuss interventions that can be designed to address the challenges of emotional labor. growing within the software engineering research community, the existing literature suggests the problem of emotional labor I. I NTRODUCTION of software developers is understudied: it has gained limited Software complexity is not solely of technological nature attention from applied psychologists working on emotional but also defined by people and processes. This means that labor due to the aforementioned stereotyping [9], [10], and special attention has to be dedicated to well-being and job has not been studied by software engineering researchers. satisfaction of people involved in creation of software. The concept of emotional labor, introduced by Hochschild II. BACKGROUND in 1983 [1], [2], refers to the “process by which workers are A. Emotional labor expected to manage their feelings in accordance with organi- zationally defined rules and guidelines”. For instance, judges Numerous studies have related emotional labor to such are expected to appear impartial, nurses—compassionate and outcomes as employee well-being, e.g., job satisfaction [3] police officers—authoritative. and burnout, as well as to organizational well-being, e.g., Software development has been traditionally stereotyped interpersonal performance and task performance [11]. as a nerdy “lone wolf” job less likely to induce emotional Morris and Feldman [12] operationalize emotional labor labor [3]. However, nowadays software developers become along four dimensions: frequency of emotional display, atten- more and more social [4] and are expected to more and tiveness to required display rules, variety of emotions required more communicate with their team mates. Moreover, indirect to be expressed and emotional dissonance, i.e., “the conflict evidence of emotional labor of software developers is abound. between genuinely felt emotions and emotions required to Already in 1991 Riedl et al. [5] reported that when debugging be displayed” [13]. In particular, emotional dissonance has experienced developers manage their emotional display, e.g., been reported to have a strong and consistent relation with by “appearing puzzled and confused, if necessary” to arouse work exhaustion and job satisfaction [14]. More recent meta- interest of their fellow developers if those might provide help analysis of 95 studies of emotional labor [15] confirmed with the debugging task. This can be seen as an example this observation and further stressed positive correlation of of surface acting, notion closely related to the emotional emotional dissonance with emotional exhaustion, depersonal- labor when “an employee changes his or her verbal, facial, ization, psychological strain, and psychosomatic complaints. and bodily expression of emotions without modifying his or Furthermore, the authors observed that surface acting corre- her underlying feelings” [6]. Furthermore, the same study of lates with the same variables. In an additional meta-analysis Riedl et al. [5] stresses that this behavior does not come study of 105 studies Kammeyer-Mueller et al. [6] concluded naturally, should be learned and not learning is experienced that stress/exhaustion levels were most substantially related to as clumsiness and a sign of lack of a novice. A more recent perceived negative display rules, i.e., perceived requirements example of an organizational rule prescribing emotional be- to suppress negative emotions, while for job satisfaction there havior is the Contributor Covenant1 the most popular code of was a substantial negative relationship with surface acting. 1 https://www.contributor-covenant.org/version/1/4/code-of-conduct.html 2 https://github.com/domgetter/NCoC 1 B. Emotional labor and software engineers skill [9], or relatively extended time periods such as work Software development has been traditionally seen as a job shifts [31], [32] or their parts [33]. The momentary approach with few interpersonal requirements [3] and, therefore, less of Gabriel and Diefendorff is exceptional in this sense [30]. likely to induce emotional labor. Not surprisingly software III. E MOTION LABOR OF SOFTWARE DEVELOPERS developers are absent from the Hochshild’s list of occupations most calling for emotional labor [1] that has influenced the In this position paper we stress that software developers are emotional labor studies in the following years [16]. subject to emotional labor, envision how emotional labor can Several studies of emotional labor of software developers be measured based on emotion detection techniques applied tend to lump them together with other kinds of IT professionals in software engineering, suggest possible antecedents and such as managers and support personnel [17], [18]. The studies consequents of emotional labor and discuss interventions that show that for the IT professionals emotional dissonance pre- can be designed to address the challenges of emotional labor. dicts work exhaustion better than traditional predictors such as A. Identification of emotions perceived workload; moreover, job satisfaction is influenced by work exhaustion and influences turnover intentions [17]. We start by discussing detection of emotions expressed by A complementary line of research focuses on software software engineers. As opposed to the existing techniques we developers: the study of Rutner et al. distinguishes between aim at the momentary detection of the emotional dissonance different job types within IT [9], and of Gunsel targeted as a gap between the emotion felt and emotion expressed. software developers [10]. Rutner et al. show that “perceptions We investigate two groups channels used to communicate of positive display rules and levels of political skill differed by expressions: biometric channels that are more likely to reflect job type, but that perceptions of negative display rules, surface emotions genuinely felt and textual channels that are more acting and deep acting did not”, justifying individual studies likely to reflect emotions required to be displayed. of software developers as opposed to other IT job types [9]. The first group of channels are physical reactions of the Furthermore, the authors state that “programmers who, like human body that can be measured by biometric devices [34]. other IT/IS professionals feel they should suppress negative Common biometric measurement techniques are electroen- emotional displays at work, also recognize the expectation to cephalography (EEG), galvanic skin response measurement express positive emotions” [9]. Gunsel has studied the relation (GSR) and measurements obtained through an eye-tracker and between emotional labor of the developers and the resulting face recognition techniques. EEG can measure valence of quality of software and observed “a positive relationship emotions, i.e., positive or negative, but also such cognitive between the variety of emotions displayed during the projects, processes as attention and perception [35]. Measuring cogni- operational effectiveness and flexibility”, while “emotional tive processes is important for understanding the impact of dissonance is found to be negatively associated with flexibility emotional labor on job satisfaction and work exhaustion as and responsiveness” [10]. they are likely to provide important confounding factors in the statistical models. Similarly, GSR can be used to measure C. Emotions and Software Engineering arousal, emotional intensity and the direction of emotion [35]. While emotional labor has rarely been studied in the context Eye tracker can be used to fatigue and relaxation. Similarly to of software engineering (Section II-B), the broader topic the cognitive processes, fatigue and relaxation can be expected of the study of emotions expressed by software engineers to affect job satisfaction and work exhaustion. Tools for has recently gained significant attention from the software emotion detection based on face recognition [36] are capable engineering research community [19], [20], [21], [22], [23], of detecting such emotions as anger, contempt, disgust, fear, [24], [25], [26], [27], [28], [29]. joy, sadness and surprise. We expect EEG, GSR and eye- tracker channels to be less regulated and therefore more ade- D. Shortcomings of the existing approaches quately representing the emotions felt by the developers. Facial The literature overview presented above suggests that there expressions are in general more regulated: non-surprisingly, is a gap between the existing studies of emotional labor, Ekman and Frisen call the face “the major nonverbal liar” [37]. on the one hand, and studies of emotions in software engi- However, since software developers work in virtual teams we neering, on the other hand, beyond the obvious differences do not expect the emotional display rules to affect the facial in the target populations. First of all, while in the software expressions. Furthermore, despite the general deceptiveness of engineering realm attention is being predominantly dedicated the facial expression micro-facial displays can provide cues to detection of natural emotions felt, e.g., by analyzing texts as to the authentic emotion felt by an individual [37]. Since created during the software development process, studies of application of biometric measurements might be error-prone emotional labor are mostly based on surveying subjects, i.e., due to their invasive character and sensor drift/noise it will be focus on the emotional labor perceived or rarely on emotions carried out in the controlled experiment setting. elicited [30]. Second, software engineering studies focus on The second group of channels are the texts produced by soft- emotions as experienced at a given moment, hic et nunc, ware developers such as code review comments, issue tracker studies of emotional labor focus on broader concepts related reports, or questions and answers on Q & A platforms. Several to personality such as dispositional affects [6] or political techniques have been proposed for detection of emotion in 2 software engineering texts: from a broad range of emotions we take leaf from the book of Gabriel and Diefendorff [30] and (anger, fear, joy, love, sadness and surprise) in the work of for continuous rating, i.e., we will record the session, replay Ortu et al. [26] to techniques focusing on detection of anger the recording to the participants and ask them to rate to what and its direction in the work of Gachechiladze et al. [27]; from extend did they feel emotional dissonance at a given moment. models of discrete emotions [26], [27] to continuous valence- Once emotional dissonance has been validated at the level of a arousal-dominance model by Mäntylä et al. [22]. As opposed single moment, one should investigate how those momentary to the emotions sensed by means of biometric instruments, values of emotional dissonance can be aggregated to extended emotions as expressed in texts are communicated towards the periods of time (cf. aggregation of software metrics [47], interlocutors, and therefore are more likely to be subject to [48]). The aggregated values can be then compared with the regulation either explicit via codes of conduct [7] or via the measurements on the emotional dissonance scale [46]. perceived notion of professional conduct. Therefore, emotions discovered by analyzing software engineering texts are more C. Antecedents and consequents of emotional labor in soft- likely to reflect emotional display rules induced by the project, ware engineering and therefore, depend on the project culture, e.g., whether it is Identified emotional labor situations should lead to un- formal or not [38], whether insults are acceptable or not [39]. derstanding antecedents and consequents of emotional Existing emotions questionnaires such as PANAS [40], labor in software engineering, i.e., aspects of developers’ LEAS [41], [42], DEQ [43] can be used for validation. personalities, roles played, project organization etc that can impact different aspects of emotional labor and the ways emo- B. From emotions to emotional labor tional labor affects software products created by developers Next we plan to relate the individual measurements obtained as well as developers’ communities. This group of activities by means of techniques developed in Section III-A to the can convert the insights obtained so far into actions that can conceptual framework of emotional labor, i.e., to propose an support software developers in their daily work. operationalization of the emotional labor constructs in 1) Antecedents: Existing studies of emotional labor an- terms of the emotions identified. tecedents concerned such as personality variables extraversion As a basis for such an operationalisation we consider the and agreeableness [3] and positive/negative affectivity [6], and four dimensions of emotional labor proposed by Morris and such demographics as gender [49], age [50], [51], race [52], Feldman [12], i.e., (1) frequency of emotional display, (2) [51] and national culture [53], [54]. All these variables can attentiveness to required display rules, (3) variety of emotions be expected to play a role also in the software engineering required to be expressed and (4) emotional dissonance. context. However, we focus on specifics of the software engi- Frequency of emotional display has been operationalised neering task and keep personality and demographic variables as frequency of interactions [12] and further proposed to as control. Specifically, we study the impact of a role played be measured as the number of interactions with different by a project contributor. Indeed, code reviewers, in particular, customers [44] or team mates [10]. Since software engineering core code reviewers [55], [56], and bug triage masters can texts have one or more addressees (e.g., individuals involved be expected to be involved in more interactions and more in reviewing a code change, or fixing a bug), ability to detect intensive interactions than regular developers. The same is emotion in software engineering envisioned in Section III-A likely to hold for project leaders and influential developers, allows one also to quantify frequency of the emotional display. closer to the center of the onion model [57], as well as for A similar argument can be made of the duration of emotional frequent contributors as opposed to the occasional ones [58]. display, one of the components of the attentiveness to required Similarly, to roles we differentiate between analytic and syn- display rules. The second component of the attentiveness to thetic software development tasks [59], e.g., identification of required display rules, i.e., intensity of the emotional display the bug cause vs. designing a bug fix, and study the impact is related to the arousal component of the valence-arousal- of the kind of the task on emotional labor. dominance model of emotions [28], [29]. Furthermore, there seems to be little attention to the impact Variety of emotions required to be expressed calls for different organisational types can have on the emotional dis- techniques capable of detecting different kinds of emotions, play rules, and, therefore, on the emotional labor. In software e.g., six discrete emotions detected by Ortu et al. [26] or the engineering, however, special attention has been given to iden- valence-arousal-dominance-based detection proposed Mäntylä tification of different organisational types, both in company- et al. [22]. Results of the variety measurement can be com- based and in open-source projects [38], [60]. We expect that pared with those obtained by the validated questionnaire [45]. distinguishing between different organisational types can help As suggested in Section III-A emotional dissonance can us to understand the differences between emotional display be seen as a discrepancy between the emotion conveyed rules induced in different software development projects. In through the biometric channels and through the textual ones. particular, we expect “community smells” [61], i.e., commu- To validate the emotional dissonance discovered in this way nication and collaboration anti-patterns reflecting undesirable we would like to build on the existing psychological scales. community characteristics such as knowledge concentration However, emotional dissonance scale proposed by Cheung and or lack of communication, to incur negative emotions on the Tang [46] is not suited for momentary evaluation. Therefore, individuals involved and might increase emotional dissonance. 3 Finally, working on different kinds of artifacts might have developers to act as culture conveyors integrating previously different impact on the emotional state of the developer. disconnected sub-communities. Furthermore, recruitment poli- Gunsel [10] has shown the system complexity has a mod- cies can be designed or adapted to select candidates with erating effect on the relation between emotional labor and self-expression congruent with emotional requirements [80]; software quality. However, since software developers’ tasks if self-expression can at least partially be detected through involve working with software artifacts one could argue that software engineering texts as suggested in Section III-A such the relation between system complexity and emotional labor is a congruence check can be integrated in the Social-Web more intricate, as reviewing or modifying more complex code candidate assessment advocated by Capiluppi et al. [81]. might not only be perceived as more intellectually challenging Finally, on a larger scale projects might consider changing but also be expected to elicit more intensive emotions. their organisational type [38], [60], e.g., by opting for a 2) Consequents: We distinguish between two groups of more/less formal communication style. consequents of interest: those related to the developers’ com- 2) Individual developer: Several emotional labor re- munity and to quality of the software produced. searchers suggested a possibility of offering trainings for First, earlier studies relating emotional labor to such conse- emotion regulation, specifically for deep acting [11]. However, quents as job satisfaction, organizational attachment/turnover there are concerns related to hidden costs of deep acting [1] intention, emotional exhaustion [3], [15], [62] should be and to differences between deep acting learned through train- replicated on software developers. Furthermore, models based ing as opposed to deep acting emerging naturally [11]. An on emotional labor should be compared against alternative alternative approach might be provided through implementa- models for turnover [63], [64], [65], [66], [67], [68], [69], tion of mindfulness techniques [82], that have been shown [70] and burnout [22], [71] designed for software engineers. to lead to significantly less emotional exhaustion and more We also plan to investigate the impact of emotional labor on job satisfaction. Application of mindfulness techniques is the software quality: preliminary results of Gunsel [10] suggest particularly promising since it has been recently successfully that such aspects of emotional labor as attentiveness, variety applied in the software engineering context as well [83]. of emotions and emotional dissonance affect software quality 3) Task: Different software development tasks can be ex- as perceived by the developers with the project complexity pected to induce different kinds of emotional labor, e.g., the moderating this relation. We would like to go beyond the triage master is likely to have a higher frequency of emotional perceived software quality to more objective measures of display. Similarly to recruitment task assignment should also software quality such as the issue fixing time (cf. the study take the risk of emotional dissonance into account, e.g., the of Ortu et al. on affectiveness vs. issue fixing time [26] and triage master should not only be technically proficient and of Jongeling et al. on the impact of the sentiment analysis aware of responsibilities of individual subteams and develop- tools in this context [72]). Furthermore, we would like to ers, but also be capable of managing the aforementioned fre- obtain a more refined understanding of the impact of emotional quent emotional display. The same argument can be made for, labor at individual activities such as introduction and removal e.g., the (core) code reviewers: one might wonder whether a of bugs, code smells and technical debt [73], [74], [75], recently observed high turnover of the core code reviewers [56] [76]. We expect this relation between emotional labor and can be attributed to emotional labor. code quality since similar relations between code quality and 4) Artifacts: Finally, if indeed as suggested in Sec- singled-out community factors have been established in the tion III-C1 maintaining or reviewing more complex systems past, e.g., socio-technical congruence [77], truck-factors [78], induces more intensive emotions, this can be used as an addi- and newcomer contributions vs. bad smells [74]. tional argument supporting efforts reducing system complexity Finally, we plan to study the impact of emotional labor such as reengineering or refactoring. on developers’ productivity. The link between emotion and IV. C ONCLUSIONS developers’ productivity has been suggested in the past [28], [22]. Our previous work covered other variables affecting pro- In this position paper we have discussed the notion of ductivity [66]: they should be included in statistical modeling. emotional labor as studied in organizational psychology and argued that emotional labor is also experienced by software D. Designing interventions developers. We have outlined the ways emotional labor can Based on the understanding of the antecedents and the be identified based on emotion detection techniques already consequents of emotional labor of software engineers one applied in software engineering, as well as suggested possible can design appropriate interventions. Interventions can take antecedents and consequents of emotional labor. Based on place at the level of the project, of the individual developer, the identified antecedents and consequents of emotional labor of their tasks and finally, at the level of the artifacts created. appropriate interventions can be designed. Some of the interventions can be supported by bots [79]. R EFERENCES 1) Project: If our expectation that community smells in- duce emotional labor then the corresponding mitigation tech- [1] A. R. Hochshild, The Managed Heart Commercialization of Human Feeling, 1st ed. University of California Press, 1983. niques identified by Tamburri et al. [61], e.g., establishing a [2] ——, The Managed Heart Commercialization of Human Feeling, 3rd ed. shared knowledge base (“social wiki”) or appointing certain University of California Press, 2012. 4 [3] J. M. Diefendorff and E. M. Richard, “Antecedents and consequences ness vs. issue fixing time,” in 12th IEEE/ACM Working Conference on of emotional display rule perceptions,” Journal of Applied Psychology, Mining Software Repositories, MSR 2015, Florence, Italy, May 16-17, vol. 88, no. 2, pp. 284–294, 2003. 2015. IEEE Computer Society, 2015, pp. 303–313. [4] M.-A. Storey, “The evolution of the social programmer,” in IEEE [27] D. Gachechiladze, F. Lanubile, N. Novielli, and A. Serebrenik, “Anger Working Conference on Mining Software Repositories, June 2012, p. and its direction in collaborative software development,” in ICSE-NIER, 140. 2017, pp. 11–14. [5] T. R. Riedl, J. S. Weitzenfeld, J. T. Freeman, G. A. Klein, and J. Musa, [28] D. Graziotin, X. Wang, and P. Abrahamsson, “Do feelings matter? on What we have learned about software engineering expertise. Berlin, the correlation of affects and the self-assessed productivity in software Heidelberg: Springer Berlin Heidelberg, 1991, pp. 261–270. engineering,” Journal of Software: Evolution and Process, vol. 27, no. 7, [6] J. D. Kammeyer-Mueller, A. L. Rubenstein, D. M. Long, M. A. Odio, pp. 467–487, 2015, jSME-13-0120.R2. B. R. Buckman, Y. Zhang, and M. D. K. Halvorsen-Ganepola, “A meta- [29] I. A. Khan, W.-P. Brinkman, and R. M. Hierons, “Do moods affect analytic structural model of dispositonal affectivity and emotional labor,” programmers’ debug performance?” Cognition, Technology & Work, Personnel Psychology, vol. 66, no. 1, pp. 47–90, 2013. vol. 13, no. 4, pp. 245–258, Nov 2011. [7] P. Tourani, B. Adams, and A. Serebrenik, “Code of conduct in open [30] A. S. Gabriel and J. M. Diefendorff, “Emotional labor dynamics: A source projects,” in IEEE 24th International Conference on Software momentary approach,” Academy of Management Journal, vol. 58, no. 6, Analysis, Evolution and Reengineering, 2017, pp. 24–33. pp. 1804–1825, 2015. [8] R. Colomo-Palacios, C. Casado-Lumbreras, S. Misra, and P. Soto- [31] B. Scott and C. Barnes, “A multilevel field investigation of emotional Acosta, “Career abandonment intentions among software workers,” labor, affect, work withdrawal, and gender,” Academy of Management Human Factors and Ergonomics in Manufacturing & Service Industries, Journal, vol. 54, no. 1, pp. 116–136, 2 2011. vol. 24, no. 6, pp. 641–655, 2014. [32] E. Ouweneel, P. M. L. Blanc, W. B. Schaufeli, and C. I. van Wijhe, [9] P. S. Rutner, F. Irani Williams, C. Campbell, and C. K. Riemenschneider, “Good morning, good day: A diary study on positive emotions, hope, “The politics of emotion: Exploring emotional labor and political skill and work engagement,” Human Relations, vol. 65, no. 9, pp. 1129–1154, across job types within the it/is profession,” SIGMIS Database, vol. 46, 2012. no. 3, pp. 52–73, Jul. 2015. [33] D. J. Beal, J. P. Trougakos, H. M. Weiss, and R. S. Dalal, “Affect [10] A. Gunsel, “The effects of emotional labor on software quality: the spin and the emotion regulation process at work,” Journal of Applied moderating role of project complexity,” Journal Of Global Strategic Psychology, vol. 98, no. 4, pp. 593–605, 2013. Management, vol. 8, no. 2, pp. 101–115, 2014. [34] A. Fountaine and B. Sharif, “Emotional awareness in software develop- [11] A. A. Grandey and A. S. Gabriel, “Emotional labor at a crossroads: ment: Theory and measurement,” in 2nd IEEE/ACM International Work- Where do we go from here?” Annual Review of Organizational Psy- shop on Emotion Awareness in Software Engineering, SEmotion@ICSE chology and Organizational Behavior, vol. 2, pp. 323–349, 2015. 2017, Buenos Aires, Argentina, May 21, 2017. IEEE Computer Society, [12] J. A. Morris and D. C. Feldman, “The dimensions, antecedents, and 2017, pp. 28–31. consequences of emotional labor,” The Academy of Management Review, [35] J. L. Andreassi, Psychophysiology: Human Behaviour and Physiological vol. 21, no. 4, pp. 986–1010, 1996. Response. Psychology Press, 1999. [13] D. R. Middleton, “Emotional style: The cultural ordering of emotions,” [36] D. McDuff, A. N. Mahmoud, M. Mavadati, M. Amr, J. Turcot, and Ethos, vol. 17, no. 2, pp. 187–201, 1989. [Online]. Available: R. E. Kaliouby, “AFFDEX SDK: A cross-platform real-time multi- http://www.jstor.org/stable/640321 face expression recognition toolkit,” in Proceedings of the 2016 CHI [14] J. A. Morris and D. C. Feldman, “Managing emotions in the Conference on Human Factors in Computing Systems, San Jose, CA, workplace,” Journal of Managerial Issues, vol. 9, no. 3, pp. 257–274, USA, May 7-12, 2016, Extended Abstracts, J. Kaye, A. Druin, C. Lampe, 1997. [Online]. Available: http://www.jstor.org/stable/40604147 D. Morris, and J. P. Hourcade, Eds. ACM, 2016, pp. 3723–3726. [15] U. R. Hülsheger and A. F. Schewe, “On the costs and benefits of [37] P. Ekman and W. V. Friesen, “Nonverbal leakage and clues to deception,” emotional labor: a meta-analysis of three decades of research,” Journal Psychiatry, vol. 32, no. 1, pp. 88–106, 1969. of Occupational Health Psychology, vol. 16, no. 3, pp. 361–389, 1 2011. [38] D. A. Tamburri, P. Lago, and H. v. Vliet, “Organizational social [16] A. S. Wharton, “The sociology of emotional labor,” Annual Review of structures for software engineering,” ACM Computing Surveys, vol. 46, Sociology, vol. 35, pp. 147–165, 2009. no. 1, pp. 3:1–3:35, Jul. 2013. [17] P. S. Rutner, B. C. Hardgrave, and D. H. McKnight, “Emotional [39] M. Squire and R. Gazda, “FLOSS as a source for profanity and insults: dissonance and the information technology professional,” MIS Quarterly, Collecting the data,” in 48th Hawaii International Conference on System vol. 32, no. 3, pp. 635–652, 2008. Sciences, HICSS 2015, Kauai, Hawaii, USA, January 5-8, 2015, T. X. [18] P. S. Rutner and C. K. Riemenschneider, “The impact of emotional Bui and R. H. S. Jr., Eds., 2015, pp. 5290–5298. labor and conflict-management style on work exhaustion of information [40] D. Watson, L. A. Clark, and A. Tellegen, “Development and validation technology professionals,” CAIS, vol. 36, p. 13, 2015. of brief measures of positive and negative affect: the PANAS scales,” [19] M. R. Wróbel, “Emotions in the software development process,” in Journal of Personality and Social Psychology, vol. 54, no. 6, pp. 1063– International Conference on Human System Interactions, June 2013, pp. 1070, 1988. 518–523. [41] R. D. Lane, D. Quinlan, G. Schwartz, P. Walker, and S. Zeitlin, “The [20] A. Kołakowska, A. Landowska, M. Szwoch, W. Szwoch, and M. R. levels of emotional awareness scale: A cognitive-developmental measure Wróbel, “Emotion recognition and its application in software engi- of emotion,” Journal of Personality Assessment, vol. 55, pp. 124–134, neering,” in 2013 6th International Conference on Human System 1990. Interactions (HSI), June 2013, pp. 532–539. [42] C. Subic-Wrana, M. E. Beutel, D. A. Garfield, and R. D. Lane, “Levels [21] E. Guzman, D. Azócar, and Y. Li, “Sentiment analysis of commit of emotional awareness: a model for conceptualizing and measuring comments in github: an empirical study,” in MSR, 2014, pp. 352–355. emotion-centered structural change,” International Journal of Psycho- [22] M. Mäntylä, B. Adams, G. Destefanis, D. Graziotin, and M. Ortu, Analysis, vol. 92, pp. 289–310, 2011. “Mining valence, arousal, and dominance: Possibilities for detecting [43] C. Harmon-Jones, B. Bastian, and E. Harmon-Jones, “The discrete burnout and productivity?” in MSR. ACM, 2016, pp. 247–258. emotions questionnaire: A new tool for measuring state self-reported [23] A. Murgia, P. Tourani, B. Adams, and M. Ortu, “Do developers feel emotions,” PLoS One, vol. 11, no. 8, pp. xx–xx, 2016. [Online]. emotions? an exploratory analysis of emotions in software artifacts,” in Available: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4976910/ Working Conference on Mining Software Repositories. New York, NY, [44] J. M. Diefendorff, M. H. Croyle, and R. H. Gosserand, “The dimen- USA: ACM, 2014, pp. 262–271. sionality and antecedents of emotional labor strategies,” Journal of [24] D. Pletea, B. Vasilescu, and A. Serebrenik, “Security and emotion: Vocational Behavior, vol. 66, no. 2, pp. 339–357, 2005. sentiment analysis of security discussions on github,” in MSR, 2014, [45] C. M. Brotheridge and R. T. Lee, “Development and validation of the pp. 348–351. emotional labour scale,” Journal of Occupational and Organizational [25] N. Novielli, F. Calefato, and F. Lanubile, “The challenges of sentiment Psychology, vol. 76, no. 3, pp. 365–379, 2003. detection in the social programmer ecosystem,” in SSE @ FSE, 2015, [46] F. Y.-l. Cheung and C. S.-k. Tang, “The effect of emotional dissonance pp. 33–40. and emotional intelligence on workfamily interference,” Canadian Jour- [26] M. Ortu, B. Adams, G. Destefanis, P. Tourani, M. Marchesi, and nal of Behavioural Science, vol. 44, no. 1, pp. 50–58, 2012. R. Tonelli, “Are bullies more productive? empirical study of affective- 5 [47] B. Vasilescu, A. Serebrenik, and M. G. J. van den Brand, “You can’t retention in FLOSS projects,” in 45th Hawaii International International control the unfamiliar: A study on the relations between aggregation Conference on Systems Science (HICSS-45 2012), Proceedings, 4-7 techniques for software metrics,” in 27th IEEE International Conference January 2012, Grand Wailea, Maui, HI, USA. IEEE Computer Society, on Software Maintenance, ser. ICSM. IEEE, 2011, pp. 313–322. 2012, pp. 3446–3455. [48] K. Mordal, N. Anquetil, J. Laval, A. Serebrenik, B. Vasilescu, and [68] M. Zhou and A. Mockus, “Who will stay in the FLOSS community? S. Ducasse, “Software quality metrics aggregation in industry,” Journal modeling participant’s initial behavior,” IEEE Trans. Software Eng., of Software: Evolution and Process, vol. 25, no. 10, 2013. vol. 41, no. 1, pp. 82–99, 2015. [49] J. Schaubroeck and J. R. Jones, “Antecedents of workplace emotional [69] B. Lin, G. Robles, and A. Serebrenik, “Developer turnover in global, in- labor dimensions and moderators of their effects on physical symptoms,” dustrial open source projects: Insights from applying survival analysis,” Journal of Organizational Behavior, vol. 21, no. 2, pp. 163–183, 2000. in 12th IEEE International Conference on Global Software Engineering, [50] J. J. Dahling and L. A. Perez, “Older worker, different actor? linking age ICGSE 2017, Buenos Aires, Argentina, May 22-23, 2017. IEEE and emotional labor strategies,” Personality and Individual Differences, Computer Society, 2017, pp. 66–75. vol. 48, no. 5, pp. 574–578, 2010. [70] M. Zhou, A. Mockus, X. Ma, L. Zhang, and H. Mei, “Inflow and [51] E. Kim, D. P. Bhave, and T. M. Glomb, “Emotion regulation in retention in OSS communities with commercial involvement: A case workgroups: The roles of demographic diversity and relational work study of three hybrid projects,” ACM Trans. Softw. Eng. Methodol., context,” Personnel Psychology, vol. 66, no. 3, pp. 613–644, 2013. vol. 25, no. 2, pp. 13:1–13:29, 2016. [52] L. Houston, “Keepin it real: race and performance ratings of positive [71] S. C. Sundaramurthy, A. G. Bardas, J. Case, X. Ou, M. Wesch, displays,” Master’s thesis, Pennsylvania State University, 12 2012. J. McHugh, and S. R. Rajagopalan, “A human capital model for [53] A. Grandey, A. Rafaeli, S. Ravid, J. Wirtz, and D. D. Steiner, “Emotion mitigating security analyst burnout,” in Eleventh Symposium On Usable display rules at work in the global service economy: the special case Privacy and Security, SOUPS 2015, Ottawa, Canada, July 22-24, 2015., of the customer,” Journal of Service Management, vol. 21, no. 3, pp. L. F. Cranor, R. Biddle, and S. Consolvo, Eds. USENIX Association, 388–412, 2010. 2015, pp. 347–359. [54] J. A. Allen, J. M. Diefendorff, and Y. Ma, “Differences in emotional [72] R. Jongeling, P. Sarkar, S. Datta, and A. Serebrenik, “On negative results labor across cultures: A comparison of chinese and u.s. service workers,” when using sentiment analysis tools for software engineering research,” Journal of Business and Psychology, vol. 29, no. 1, pp. 21–35, Mar 2014. Empirical Software Engineering, Jan 2017. [55] A. Bacchelli and C. Bird, “Expectations, outcomes, and challenges of [73] A. F. Yamashita and L. Moonen, “Do developers care about code modern code review,” in ICSE. IEEE Press, 2013, pp. 712–721. smells? an exploratory survey,” in 20th Working Conference on Reverse [56] P. van Wesel, B. Lin, G. Robles, and A. Serebrenik, “Reviewing career Engineering, WCRE 2013, Koblenz, Germany, October 14-17, 2013, paths of the openstack developers,” in ICSME, 2017. R. Lämmel, R. Oliveto, and R. Robbes, Eds. IEEE Computer Society, [57] K. Nakakoji, Y. Yamamoto, Y. Nishinaka, K. Kishida, and Y. Ye, 2013, pp. 242–251. “Evolution patterns of open-source software systems and communities,” [74] M. Tufano, F. Palomba, G. Bavota, R. Oliveto, M. Di Penta, A. De in IWPSE, 2002, pp. 76–85. Lucia, and D. Poshyvanyk, “When and why your code starts to smell [58] B. Vasilescu, A. Serebrenik, M. Goeminne, and T. Mens, “On the bad (and whether the smells go away),” IEEE Transactions on Software variation and specialisation of workload – a case study of the Gnome Engineering, vol. PP, no. 99, pp. 1–1, 2017. ecosystem community,” Empirical Software Engineering, vol. 19, no. 4, [75] G. Bavota and B. Russo, “A large-scale empirical study on self-admitted pp. 955–1008, 2013. technical debt,” in Proceedings of the 13th International Conference on [59] G. Schreiber, H. Akkermans, A. Anjewierden, R. de Hoog, N. Shad- Mining Software Repositories, MSR 2016, Austin, TX, USA, May 14-22, bolt, W. Van de Velde, and B. Wielinga, Knowledge Engineering and 2016, M. Kim, R. Robbes, and C. Bird, Eds. ACM, 2016, pp. 315–326. Management: The CommonKADS Methodology. Cambridge, MA: MIT [76] E. Maldonado, R. Abdalkareem, E. Shihab, and A. Serebrenik, “An Press, 1999. empirical study on the removal of self-admitted technical debt,” in Pro- [60] D. A. Tamburri, P. Lago, and H. van Vliet, “Uncovering latent social ceedings of the 33rd International Conference on Software Maintenance communities in software development,” IEEE Software, vol. 30, no. 1, and Evolution, ICSME 2017, Shanghai, China, September 20-22, 217. pp. 29–36, Jan 2013. IEEE, 2017, pp. xx–xx. [61] D. A. Tamburri, P. Kruchten, P. Lago, and H. v. Vliet, “Social debt [77] I. Kwan, A. Schröter, and D. Damian, “Does socio-technical congruence in software engineering: insights from industry,” Journal of Internet have an effect on software build success? A study of coordination in a Services and Applications, vol. 6, no. 1, p. 10, May 2015. software project,” IEEE Trans. Software Eng., vol. 37, no. 3, pp. 307– [62] N. J. Yanchus, L. T. Eby, C. E. Lance, and S. Drollinger, “The impact 324, 2011. of emotional labor on work–family outcomes,” Journal of Vocational [78] G. Avelino, L. T. Passos, A. C. Hora, and M. T. Valente, “A novel Behavior, vol. 76, no. 1, pp. 105–117, 2010. approach for estimating truck factors,” in 24th IEEE International [63] E. Constantinou and T. Mens, “Socio-technical evolution of the ruby Conference on Program Comprehension, ICPC 2016, Austin, TX, USA, ecosystem in github,” in IEEE 24th International Conference on Soft- May 16-17, 2016. IEEE Computer Society, 2016, pp. 1–10. ware Analysis, Evolution and Reengineering, SANER 2017, Klagenfurt, [79] M.-A. Storey and A. Zagalsky, “Disrupting developer productivity one Austria, February 20-24, 2017, M. Pinzger, G. Bavota, and A. Marcus, bot at a time,” in Proceedings of the 2016 24th ACM SIGSOFT Eds. IEEE Computer Society, 2017, pp. 34–44. International Symposium on Foundations of Software Engineering, ser. [64] V. Midha and P. Palvia, “Retention and quality in open source software FSE 2016. New York, NY, USA: ACM, 2016, pp. 928–931. projects,” in Reaching New Heights. 13th Americas Conference on [80] R. W. Arvey, G. L. Renz, and T. W. Watson, “Emotionality and job Information Systems, AMCIS 2007, Keystone, Colorado, USA, August performance: Implications for personnel selection.” Research in person- 9-12, 2007, J. A. Hoxmeier and S. Hayne, Eds. Association for nel and human resources management : a research annual, vol. 16, pp. Information Systems, 2007, p. 25. 103–148, 1998. [65] A. Schilling, “What do we know about FLOSS developers’ attraction, [81] A. Capiluppi, A. Serebrenik, and L. Singer, “Assessing technical candi- retention, and commitment? A literature review,” in 47th Hawaii Inter- dates on the social web,” IEEE Software, vol. 30, no. 1, pp. 45–51, Jan national Conference on System Sciences, HICSS 2014, Waikoloa, HI, 2013. USA, January 6-9, 2014. IEEE Computer Society, 2014, pp. 4003– [82] U. R. Hülsheger, H. J. E. M. Alberts, A. Feinholdt, and J. W. B. 4012. Lang, “Benefits of mindfulness at work : the role of mindfulness in [66] B. Vasilescu, D. Posnett, B. Ray, M. G. J. van den Brand, A. Serebrenik, emotion regulation, emotional exhaustion, and job satisfaction,” Journal P. Devanbu, and V. Filkov, “Gender and tenure diversity in GitHub of Applied Psychology, vol. 98, no. 2, pp. 310–325, 1 2013. teams,” in CHI Conference on Human Factors in Computing Systems, [83] B. Bernárdez, A. Durán, J. A. Parejo, and A. Ruiz-Cortés, “An ex- ser. CHI. ACM, 2015, pp. 3789–3798. perimental replication on the effect of the practice of mindfulness in [67] A. Schilling, S. Laumer, and T. Weitzel, “Who will remain? an eval- conceptual modeling performance,” Journal of Systems and Software, uation of actual person-job and person-team fit to predict developer 2016. 6