Retrofitting Meetings for Psychological Safety Marios Constantinidesa , Sagar Joglekara and Daniele Querciaa,b a Nokia Bell Labs, Cambridge, UK b King’s College, London, UK Abstract Meetings are the fuel of organizations’ productivity. At times, however, they are perceived as wasteful vaccums that deplete employee morale and productivity. Current meeting tools, to a great extent, have simplified and augmented the ways meetings are conducted by enabling participants to “get things done” and experience a comfortable physical environment. However, an important yet less explored element of these tools’ design space is that of psychological safety—the extent to which participants feel listened to, or motivated to be part of a meeting. We argue that an interdisciplinary approach would benefit the creation of new tools designed for retrofitting meetings for psychological safety. This approach comes with not only research opportunities—ranging from sensing to modeling to user interface design—but also challenges—ranging from privacy to workplace surveillance. Keywords meetings, execution, physical comfort, psychological safety 1. Introduction port them. While meeting tools pledge to fa- cilitate better meetings—ones that are well- Meetings are often considered as the fuel of executed, and create a safe environment for an organization’s productivity. Employees contribution—yet report after report estimate come together for a common purpose to dis- a growth of ineffective meetings1 ; numbers cuss ideas, to make collective decisions, and though that are bound to change, if meeting to ultimately reach their objectives. How- tools were to fully capture people’s meetings ever, it is no secret that meetings are of- experience. Recently, in a large-scale crowd- ten seen as wasteful vacuums, or as an en- sourcing study [2], researchers determined emy of productivity. Although there are the main factors that impact people’s meet- good meetings and bad meetings, their col- ings experience. They administered a 28- lective negative impact on employee morale item questionnaire to 363 individuals whose and productivity is significant [1]. To moder- answers were statistically analyzed through ate this, organizations devote notably large Principal Component Analysis, and found amounts of resources to facilitate and sup- that three factors sufficiently capture peo- ple’s experience in meetings, namely, (a) exe- Joint Proceedings of the ACM IUI 2021 Workshops, April cution, (b) physical comfort, and (c) psycho- 13–17, 2021, College Station, USA logical safety. Put differently, these factors " marios.constantinides@nokia-bell-labs.com (M. Constantinides); sagar.joglekar@nokia-bell-labs.com (S. capture the extent to which (a) people feel Joglekar); daniele.quercia@nokia-bell-labs.com (D. that a meeting was productive, (b) the meet- Quercia) ing room was pleasant, and (c) the setting  0000-0003-1454-0641 (M. Constantinides); was psychologically safe. 0000-0002-8388-9137 (S. Joglekar); 0000-0001-9461-5804 (D. Quercia) © 2021 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 1 https://www.forbes.com/sites/danielreed/2019/01/ 4.0 International (CC BY 4.0). CEUR Workshop Proceedings 30/report-suggests-that-23rds-of-the-100-billion-spent -annually-on-business-meetings-travel-is-wasted CEUR http://ceur-ws.org (CEUR-WS.org) Workshop ISSN 1613-0073 Proceedings Current meeting tools, however, primarily 2. Capturing psychological focus on enabling participants to “get things done” (i.e., execution), or on optimizing the safety environmental conditions (i.e., physical com- Psychological safety is about whether par- fort). Execution is about whether a meet- ticipants felt listened to, or motivated to be ing had a clear structure, purpose, and re- part of a meeting. Edmondson described it sulted into a list of actionable items; a large “as the absence of interpersonal fear that al- body of previous research focused on these lows people to speak up with work-relevant aspects. Kim and Rudin [3] developed a sys- content” [11]. Previous research showed that tem that detects key decisions in dialogues, inclusiveness and balance of conversational while McGregor and Tang [4]’s system gen- turn-taking play an important role in group erates an ‘action items’ list from the spoken performance [12]. Tools have been devel- dialogue. Using an agenda planning tech- oped to create awareness by highlighting nique, Garcia et al. [5] allowed meetings par- salient moments during conversations [13], ticipants to vote for agendas to improve per- to enhance group collaboration through per- ceived meeting quality. Video Threads [6] suasive feedback [14], and to allow partici- provides asynchronous video sharing, while pants to reflect on their own and their peers’ Time Travel Proxy [7] identifies the gist of experiences [15]. Although such tools, to what was missed to enable late participa- some extent, offer features that facilitate ex- tion effectively. Physical comfort is about ecution and support physical comfort, they whether the meeting room was pleasant (in often fall short in enabling psychological regard to air quality and crowdedness). In safety. We argue that a more interdisci- Human-Building Interaction literature [8], plinary approach would likely benefit the poor environmental conditions are known to creation of new tools designed for support- impact employees’ cognitive functions, de- ing psychological safety. To achieve that, we cision making, and performance. There- foresee a number of challenges and opportu- fore, organizations have resorted to sensors nities, ranging from sensing to modeling to through which the environmental conditions user interface (UI) design. could be sensed [2, 9], and even adapted ac- Sensing. New emerging sensing devices cordingly [10] to meet recommended stan- such as earables [16] are now fully equipped dards, thus increasing their employees’ pro- with IMU (inertial measurement unit) sen- ductivity and well-being. To this end, meet- sors, allowing on-body and on-device sens- ing tools could account for physical comfort ing. This opens up a new avenue for meet- in their design, potentially in a form of inter- ing tools by allowing participants to mon- ventions (e.g., adjusting ventilation). How- itor signals that could otherwise go un- ever, the design space should not only facilitate noticed; for example, capturing body cues execution and ensure that desirable physical of (dis)agreement or (in)active participa- environmental conditions are met, but should tion during a virtual conversation when the also cultivate a psychologically safe setting. video stream is absent [17]. Similarly, smart- watches are now fully equipped with heart rate sensors that provide a window to peo- ple’s physiology, allowing one to track their own or their peers’ emotional states [18]; as- pects that are closely linked to creating a safe environment for contribution. Additionally, and empathic of each other [24]. Borrow- in the future, we foresee that better precision ing ideas from calm technology [25] and bio- devices would enable more nuanced non- philic design, meeting tools could embrace verbal or verbal communication patterns to new types of cues only available through be captured. technology [26]. For example, the use of dif- Modeling. New algorithms are also ferent symbols, imagery, and artificial arti- likely to provide a new understanding and facts (e.g., real-world objects [27], light [28], perspective that would help further theorize or movement 2 ) could augment the ways we the concept of psychological safety. New interact and communicate with each other. Natural Language Processing text-mining These new visualizations could bring teams algorithms [19] are now able to reveal certain together despite working apart, and remove language markers that might be deeply any geographical barriers due to physical dis- hidden in a conversation, particularly in tancing. a remote setting. These algorithms can Workplace surveillance. While this in- annotate everyday language and capture terdisciplinary approach promises to deliver important types of social interactions (e.g., experiences richer of psychological safety, a heated discussion resulting in conflict it also raises questions relating to work- resolution). NLP-based algorithms can now place surveillance. It is often regarded that analyze conversations and test whether these organizations and surveillance go hand in conversations accommodate different points hand [29]. On a very pragmatic level, there of view [20], or even reveal the presence (or is a handful of reasons as to why organi- absence) of certain health-related markers zations opt in for employees’ surveillance (e.g., stress markers) [21]. Additionally, new (e.g. maintaining productivity, monitoring Natural Sound Processing algorithms [22] resources used, protecting the organization are now able to model verbal cues (e.g., from legal liabilities). The critics, however, prosody) that would potentially enable richer rightly argue that there is a fine line between and more focused interactions. For example, what organizations could be monitoring and prosodic features (e.g., pitch and energy) what they should be monitoring. If crossed, it are known to provide a reliable indication will have consequences on employees, affect- of the emotional status in a conversational ing their well-being, work culture, and pro- exchange. ductivity. If future meetings tools incorpo- UI design. Last but not least, new oppor- rate any kind of employees’ monitoring, they tunities are likely to arise for the UI design need to ensure that is done in a way that pre- community. New visualizations are more serves an individual’s rights, including that likely to be (re)invented, beyond dashboards of privacy. and simple analytics [23]. Drawing from be- The workplace is constantly changing havioral economics research, we foresee that and evolving. These changes are currently new forms of interventions would allow peo- accelerated by the COVID-19 pandemic, ple to be more empathetic, compassionate, which might be leading to a dramatic change and aware of each other’s emotional states, in how we work: from the well-known views, and thoughts. Previous research in the eight-hour workday to the office building area of organizational behavior showed that to the salient boundaries between work and affective sharing within groups conveys our personal life. Meetings are no exception internal experiences, signals our emotional states and, potentially, makes us more aware 2 Blooming: http://www.thelisapark.com/blooming to this sudden change. In a post-pandemic ative Work, volume 6, ACM, 2004, pp. world, we envision that new sensing devices 346–349. would provide access to employees’ data [6] J. Barksdale, K. Inkpen, M. Czerwin- that otherwise might not be possible to ski, A. Hoff, P. Johns, A. Roseway, collect (e.g., on-body sensing); that new G. Venolia, Video threads: Asyn- algorithms would ‘make sense’ of such chronous video sharing for temporally data, and capture behavioral aspects that distributed teams, in: Proc. of the ACM are hard to quantify (e.g., (dis)agreement, 2012 Conference on Computer Sup- empathy, or stress markers); and that new ported Cooperative Work, ACM, 2012, user interfaces (e.g., inspired by biophilic pp. 1101–1104. design) would enable meeting participants [7] J. Tang, J. Marlow, A. Hoff, A. Roseway, to stay connected despite any geographical K. Inkpen, C. Zhao, X. Cao, Time travel or technological barriers due to remote proxy: Using lightweight video record- working. ings to create asynchronous, interactive meetings, in: Proc. of the ACM CHI Conference on Human Factors in Com- References puting Systems, ACM, 2012, pp. 3111– 3120. [1] S. Kauffeld, N. Lehmann-Willenbrock, [8] H. S. Alavi, E. F. Churchill, M. Wiberg, Meetings matter: Effects of team meet- D. Lalanne, P. Dalsgaard, A. Fatah gen ings on team and organizational suc- Schieck, Y. Rogers, Introduction to cess, Small Group Research 43 (2012) human-building interaction (hbi) inter- 130–158. facing hci with architecture and urban [2] M. Constantinides, S. Šćepanović, design, ACM Trans. on Computer- D. Quercia, H. Li, U. Sassi, M. Eggle- Human Interaction 26 (2019). ston, Comfeel: Productivity is a matter [9] H. S. Alavi, H. Verma, M. Papinutto, of the senses too, Proc. of the ACM D. Lalanne, Comfort: A coordi- on Interactive, Mobile, Wearable and nate of user experience in interactive Ubiquitous Technologies 4 (2020) 1–21. built environments, in: IFIP Confer- [3] B. Kim, C. Rudin, Learning about ence on Human-Computer Interaction, meetings, Data Mining and Knowledge Springer, 2017, pp. 247–257. Discovery 28 (2014) 1134–1157. [10] P. Bader, A. Voit, H. V. Le, P. W. Woź- [4] M. McGregor, J. C. Tang, More to niak, N. Henze, A. Schmidt, Win- meetings: Challenges in using speech- dowwall: Towards adaptive buildings based technology to support meetings, with interactive windows as ubiquitous in: Proc. of the ACM Conference displays, ACM Trans. on Computer- on Computer Supported Cooperative Human Interaction 26 (2019) 1–42. Work and Social Computing, ACM, [11] A. Edmondson, Psychological safety 2017, pp. 2208–2220. and learning behavior in work teams, [5] A. B. Garcia, J. Kunz, M. Fischer, Cut- Administrative Science Quarterly 44 ting to the chase: Improving meet- (1999) 350–383. ing effectiveness by focusing on the [12] A. W. Woolley, C. F. Chabris, A. Pent- agenda, in: Proc. of the ACM Confer- land, N. Hashmi, T. W. Malone, Evi- ence on Computer Supported Cooper- dence for a collective intelligence factor in the performance of human groups, Science 330 (2010) 686–688. volume 13, 2019, pp. 428–439. [13] T. Bergstrom, K. Karahalios, Conver- [21] S. Scepanovic, E. Martin-Lopez, sation clusters: Grouping conversation D. Quercia, K. Baykaner, Extracting topics through human-computer dia- medical entities from social media, log, in: Proc. of the ACM CHI Confer- in: Proc. of the ACM Conference on ence on Human Factors in Computing Health, Inference, and Learning, 2020, Systems, ACM, 2009, pp. 2349–2352. pp. 170–181. [14] T. Kim, A. Chang, L. Holland, A. S. Pent- [22] K. Curtis, G. J. Jones, N. Campbell, land, Meeting mediator: Enhancing Effects of good speaking techniques on group collaboration using sociometric audience engagement, in: Proc. of feedback, in: Proc. of the ACM Confer- the ACM International Conference on ence on Computer Supported Coopera- Multimodal Interaction, 2015, pp. 35– tive Work and Social Computing, ACM, 42. 2008, pp. 457–466. [23] S. Few, Information Dashboard Design: [15] B. A. Aseniero, M. Constantinides, The Effective Visual communication of S. Joglekar, K. Zhou, D. Quercia, data, O’Reilly Media, 2006. Meetcues: Supporting online meetings [24] F. Walter, H. Bruch, The positive group experience, in: Proc. of the IEEE Vi- affect spiral: A dynamic model of the sualization Conference, IEEE, 2020, pp. emergence of positive affective similar- 236–240. ity in work groups, Organizational Be- [16] F. Kawsar, C. Min, A. Mathur, A. Mon- havior 29 (2008) 239–261. tanari, Earables for personal-scale be- [25] M. Weiser, J. S. Brown, The coming havior analytics, IEEE Pervasive Com- age of calm technology, in: Beyond puting 17 (2018) 83–89. calculation, Springer, 1997, pp. 75–85. [17] J.-H. Choi, M. Constantinides, [26] C. Y. Qin, M. Constantinides, L. M. S. Joglekar, D. Quercia, Kairos: Aiello, D. Quercia, Heartbees: Visualiz- Talking heads and moving bodies for ing crowd affects, in: Proc. of the IEEE successful meetings, in: Proc. of the VIS Arts Program (VISAP), IEEE, 2020, International Workshop on Mobile pp. 1–8. Computing Systems and Applications [27] B. Yu, M. Funk, J. Hu, L. Feijs, Stresstree: (HotMobile), 2021, pp. 1–7. A metaphorical visualization for [18] S. Park, M. Constantinides, L. M. Aiello, biofeedback-assisted stress manage- D. Quercia, P. Van Gent, Wellbeat: ment, in: Proc. of the Conference on A framework for tracking daily well- Designing Interactive Systems, 2017, being using smartwatches, IEEE pp. 333–337. Internet Computing 24 (2020) 10–17. [28] B. Yu, J. Hu, M. Funk, L. Feijs, Delight: [19] M. Choi, L. M. Aiello, K. Z. Varga, Biofeedback through ambient light for D. Quercia, Ten social dimensions of stress intervention and relaxation assis- conversations and relationships, in: tance, Personal and Ubiquitous Com- Proc. of The Web Conference (WWW), puting 22 (2018) 787–805. 2020, pp. 1514–1525. [29] K. Ball, Workplace surveillance: An [20] A. Robertson, L. M. Aiello, D. Quercia, overview, Labor History 51 (2010) 87– The language of dialogue is complex, in: 106. Proceedings of the International AAAI Conference on Web and Social Media,