Tangible interventions for office work well-being: approaches, classification, and design considerations

O ﬃ ce well-being aims to explore and support a healthy, balanced and active work style in o ﬃ ce environments. Recent work on tangible user interfaces has started to explore the role of physical, tangible interfaces as active interventions to explore how to tackle problems such as inactive work and lifestyles, and increasingly sedentary behaviours. We identify a fragmented research landscape on tangible O ﬃ ce well-being interventions, missing the relationship between interventions, data, design strategies, and outcomes, and behaviour change techniques. Based on the analysis of 40 papers, we identify 7 classi ﬁ cations in tangible O ﬃ ce well-being interventions and analyse the intervention based on their role and foundation in behaviour change. Based on the analysis, we present design considerations for the development of future tangible O ﬃ ce well-being design interventions and present an overview of the current ﬁ eld and future research into tangible O ﬃ ce well-being interventions to design for a healthier and active o ﬃ ce environment.


Introduction
In recent years, we observe an emergence in technology that explores how to support or mediate 'Office wellbeing'a healthy, balanced and active work style in office environments (Damen et al. 2020a). These technologies have the potential to address the increasing group of individuals who, because of their work, have inactive work and lifestyles, and are increasingly sedentary (Åborg and Billing 2003;Clemes, O'Connell, and Edwardson 2014;Parry and Straker 2013;Waters et al. 2016). The risks of having a non-active and/or sedentary lifestyle are associated with non-communicable diseases such as type II diabetes and cardiovascular diseases (Biddle 2007;Dias, Vianna, and Barbosa 2022;Gibbs et al. 2015;Healy et al. 2008;Schroeder 2007). Office work has even further evolved during the Covid pandemic where the environment has changed into a hybrid setting where people change their work location to both the office as well as the work-life balance of individuals (Wang et al. 2021). Numerous design interventions have been proposed in recent years, with a dominant focus on the reduction of sedentary and inactive behaviour in the office environment (for an overview see: Huang, Benford, and Blake 2019 and Damen et al. 2020a). These interventions have predominantly materialised in the form of digital, non-tangible 'apps' on phones or wearables (Huang, Benford, and Blake 2019) to measure and visualise the health, (work) behaviour, or activity of individuals (Brakenridge et al. 2016;Dallinga et al. 2018;Jin et al. 2022;Renner et al. 2020;Sullivan and Lachman 2017). While 'apps' are the dominant way of interacting with personal health or activity data, prior work demonstrated its limitations, including 'display blindness' (Müller et al. 2009), attention overload through notifications (Shirazi et al. 2014), low recall in content (Müller et al. 2009;Pearson 2021) or lack of social and contextual situated information (Bakker, van den Hoven, and Eggen 2015;Brombacher et al. 2019). Next to this, there is a dearth of research on interventions using connected devices others than smartphones, and see an opportunity for more tangible focused designs developed for users in their context (Huang, Benford, and Blake 2019). Therefore, a large body of research has started to explore how tangible or physical computing approaches can be prioritised over the app model, thus, enabling newly situated contextually relevant interventions within the office environment (Huang, Benford, and Blake 2019). Inspired by research in Tangible User Interfaces (TUI) (Ishii and Ullmer 1997;Shaer and Hornecker 2009), recent work has started to explore the role of tangible interfaces as active interventions in Office wellbeing. This work leverages prior findings in physical and tangible computing (such as Buur 2006 andShaer andHornecker 2009) to mediate and support active ways of working in office environments that go beyond the use of an app. Tangible user interfaces have been shown to enhance legibility due to their open visibility, increased engagement, support social awareness, social collaboration, and interaction and provide an external record of previous states and actions (Shaer and Hornecker 2009). Tangible user interfaces have also been shown to have a more inviting quality than digital and screen-based interventions. These artifacts are not just residing on a screen, but are part of the actual physical environment (Shaer and Hornecker 2009). Tangible approaches have taken a diverse range of varying perspectives including how to support behaviour change, targeted behaviour, and technology type (type and input measure) (Damen et al. 2020a). The goal of these tangible interfaces is to embed interventions into everyday life and work practice to support feedback and visualisation on Office well-being in situ and context. While the mentioned work on tangible interfaces opens a new interesting approach to how to affect Office well-being, (i) the diversity of design approaches, (ii) the use of definitions and concepts, and (iii) the relation to the behaviour change are not well understood and are fragmented (Constantinides et al. 2020;Huang, Benford, and Blake 2019;Orji and Moffatt 2018). This spectrum of Office well-being design tangible interventions has led to a somewhat fragmented research landscape, missing the relationship between interventions, data, design strategies, outcomes, and behaviour change techniques. This gap opens up a necessity to explore and better define the scope, unit of analysis, and approaches used in tangible Office well-being interventions, including emphasising the importance of a comprehensive understanding of Office well-being interventions for supporting relevant design research and practices.
We discuss the domain to clarify the focus and scope of tangible Office well-being interventions research and reflect on the characterisation and properties of such tangibles. Based on an analysis of recent papers from the domain of tangible Office well-being interventions, we discuss the classification that characterises tangible interventions in a number of dimensions such as: input, output, time, form, spatial relation, target behaviour, and user interaction. Next, we analyse these artifacts based on their behaviour change techniques.
Based on this classification, we reflect on the opportunities and challenges and give design considerations for the development of future tangible Office wellbeing interventions. Researchers can use this work as a starting point for future research into tangible Office well-being interventions to design for a healthier and more active work environment.

Office well-being interventions
In recent years, many academic disciplines (including Human-Computing Interaction, HCI) have been concerned with the topic of well-being in the workplace. Several definitions, classifications, and application areas have been opted to identify physical activity interventions. These include the application areas for physical activity interventions (Consolvo et al. 2006), types of technologies to promote physical activity (Peeters and Megens 2014), types of technologies to characterise physical activity-focused interventions (Ren 2019), and the Behavioural intervention technology model (Mohr et al. 2014). The models and application areas above show a widespread of classification for artifacts, not specifically for the office environment, and often miss a user-centered design approach where users, tasks, and environments are not considered within the identification of physical activity interventions. wellbeing covers both psychological and physical components (such as physical activity) and is considered a crucial aspect of productivity, employability, and sustainable work performance (de Jonge and Peeters 2019; Sandblad et al. 2003;van Scheppingen et al. 2014). A few decades ago, employee well-being in the workplace, including aspects such as job satisfaction and job commitment (Iaffaldano and Muchinsky 1985) has received growing awareness in the field of work and occupational psychology-related research (see e.g. Dorenbosch 2009). Scholars have worked on understanding factors of the degree of alignment between the individual and their (work) environment (Person-Environment fit), by focusing on the alignment with the organisation, the job, the co-workers, and the supervisor (Edwards 2008). At the same time, enhancing physical activity and reducing sedentary behaviour has become an important research issue in the field of work and (public) health (see e.g. Chau et al. 2010).
Nowadays, well-being has become a top priority in the work setting, especially in desk-based occupations, for a variety of reasons, at a micro level (i.e. the individual employee | e.g. stress, physical inactivity), at a meso (i.e. the employer | e.g. reduced performance, absenteeism, retention of talent) and at a macro level (i.e. society | e.g. health care costs). Also, more emphasis is put on the role of technology in Office well-being. Technology has shaped the way we work and has modified the work itself. Technological developments have enabled a high level of communication via smartphones, tablets, laptops, video conferencing systems, etcetera. Combined with fast internet this results in faster processes and allows office workers to be accessible and connected at anytime and anywhere. Hence, the boundaries between family life, work, and personal life are blurring, resulting in negative effects such as work-home conflicts, information overload, physical inactivity, violation of privacy, stress, and burnout (Peeters and Megens 2014). While, on a work level services such as Email, Skype, and remote company data points have provided employees with all the needed information at a single location, without any need to be physically active (de Jonge and Peeters 2019).

Approach and methodology
The goal of this paper is to explore and define the field of tangible Office well-being interventions. Due to the fragmented research landscape, it is currently not clear what the relationships are between interventions, data, design strategies, and outcomes and behaviour change techniques (Damen et al. 2020a;Huang, Benford, and Blake 2019;Orji and Moffatt 2018). With this paper, we aim to bring conceptual clarity to the topic of tangible Office well-being and propose a new classification to categorise and describe relevant work in one consistent frame of reference. The methodology of our work is twofold: (i) we created a corpus of 40 key papers that are identified as 'tangible Office well-being interventions', and (ii) we analyse and code the corpus to deduct a classification that characterises the various approaches within 'tangible Office well-being'. Using this methodology, we finally present a set of design considerations that elucidate various trade-offices and approaches for designing future 'tangible Office well-being'. The 40 design papers consist of 24 full papers (full conference papers and journals) and 16 short papers (case studies, demonstrations, provocations, pictorials, late-breaking work, extended abstract).
The first step in our methodology was to select related literature by collecting a set of relevant papers that self-identified as being related to 'tangible Office well-being'. This corpus is not meant as an exhaustive systematic literature, but to collate a set of relevant and representative papers that enable us to conduct a meta-analysis of their design approaches. Due to the fragmented landscape of interventions, an open-ended approach and keyword approach was taken (Elliott and Timulak 2015) to combine the strengths of both approaches, using the open-ended approach to generate a broad and explorative corpus. The keyword-based approach was used to create a clear cut-off for the selected tangible Office well-being interventions. This led to a final corpus of 40 papers that we believe is an accurate selection of representative papers for this topic.
The second step in our methodology was to code and analyse the corpus of 40 papers to derive a classification. The analysis resulted in several classifications including the role of data, output mechanisms, form, space relation, target behaviour, and user interaction. This classification scheme was inspired by common topics in HCI design spaces (Bressa et al. 2021;Damen et al. 2020a;Kitson, Prpa, and Riecke 2018;Ledo et al. 2018) including space, time, input, output, data and technology. We extended our search by curating and selecting additional case studies to further expand the scheme, while providing contrasting perspectives on these themes to create a broader categorisation of tangible Office well-being intervention and identify possible new directions for this emerging topic. We then analysed the entire corpus to refine and finalise the design classification schemes. Using this final classification, we present design considerations for future work.

Analysis of 'tangible Office well-being' literature
We analysed a set of representative papers from different research communities based on keyword search to create an overview of selected literature on 'tangible Office well-being.

Criteria, keywords, and analysis
The research field was analysed, using scoping review approach, to clarify the focus and scope of tangible 'Office well-being intervention' research and reflect on the characterisation and properties of such tangibles. We collected 40 papers (Figure 1) with the following inclusion criteria focusing on tangible Office wellbeing interventions. We started with the same search strategy used by Damen et al. (Damen et al. 2020a), looking specifically at tangible designs. The review was re-analysed and the tangible interventions of this review were selected. Damen et al., selected 45 artifacts in their scoping review published between 2009 and 2019, targeting physical activity and/or sedentary behaviour, being used (partly) during office hours and including digital technology in their delivery. From this review, we excluded the 26 interventions due to them being purely digital (mostly being applications) and the dearth of research interventions other than smartphones (Huang, Benford, and Blake 2019). After excluding the purely digital interventions, 19 artifacts were included in the analysis. To extend the review, we repeated the search including the years 2020 and 2021, resulting in 6 additional artifacts. Additionally, a keyword search was conducted in the ACM library, Scopus, and IEEE Xplore liberty. The keywords included 'sedentary behaviour', 'sitting', 'physical activity', 'inactivity' 'office' and 'work', because are strong indicators for work well-being. To also include papers from a broader perspective on healthy office behaviours, an additional keyword search was conducted focusing on office wellbeing. This search included keywords: 'productivity', 'posture', 'work wellbeing', 'posture change', 'office work breaks' and 'office wellbeing'. This search led to an additional 15 artifacts, creating a total of 40 artifacts included in the analysis of the field of Office well-being.
We included full papers, journals, work-in-progress, late-breaking work, demonstrators, and short papers in our search. Papers were not excluded based on their venue, leading to a broad range of venues including CHI, DIS, TEI, OZCHI, ISWC, and Ubicomp. Papers were included if they were published between 2009 and 2021. Papers were excluded if they did not include an artifact, were purely digital (e.g. applications), focused on children and/or patients, or were not designed for the office environment.

Analysis
The design interventions were analysed in three steps. First, an initial open-coding scheme (Charmaz 2014) was developed by one of the authors based on selected case studies on tangible Office well-being interventions. Second, the coding scheme was discussed and refined with two co-authors. Third, all artifacts were defined in the coding scheme, while refining and revisiting the classification. After a final revision, the final classification scheme was set based on 7 classifications: (i) input, (ii) output, (iii) time, (iv) form, (v) space relation, (vi) target behaviour, and (vii) user interaction. Based on the 7 classifications, the artifacts were re-analysed leading to the final classification of the designs ( Table 2). The sub-categories in the classifications (input, output, and space relation) were set after a discussion with all authors, based on additional analysis of the design interventions. Additionally, we looked at the grounding of the intervention in behaviour change theory. The Transtheoretical Model (Prochaska, Johnson, and Lee 2009) was used to learn which of the stages of change the interventions would fit (Table 1). This model is used due to it being the most commonly used model in Office well-being interventions. (Damen et al. 2020a;Huang, Benford, and Blake 2019;Orji and Moffatt 2018) and fits different levels of readiness, fitting interventions that are developed at different stages of change and can be used for practitioners or researchers for different fields including clinical and public health interventions (Nigg et al. 2011). The artifacts are classified based on the used behaviour change techniques in the artifact. The used behaviour change techniques were linked to the processes of change, which were linked to the stage of change of the artifact.

Classification of the tangible Office wellbeing field
To reflect on the use and characterisation of 'Tangible Office well-being' interventions, we reflect on the field of Office well-being that was derived from analyzing the design approaches of 40 interventions collated in our corpus. We discuss and exemplify 7 key design classifications: (i) input, (ii) output, (iii) time, (iv) form, (v) space relation, (vi) target behaviour, and (vii) user interaction. Additionally, we describe an initial mapping on how the work in our corpus connects to a theocratical grounding of behaviour change.

Data input
Data input is divided into three subcategories (i) focusing on the data itself (steps, heart rate, posture, user presence/movement, light, sound, air quality, productivity, and others ( Figure 2)), (ii) the source of the data (personal, environmental or building), and (iii) the relation to the data (individual or social/group related data). From a data perspective, interventions most frequently use personal data (ƒ = 30/40, 75%). Personal data focuses on sedentary behaviour where the artifact measures if individuals are present and/or moving (ƒ = 10/40, 25%), but also their sitting time behind their office desk (ƒ = 5/40, 12,5%) or their sitting posture (ƒ = 1/40, 4%). This sedentary behaviour is measured in a variety of ways with sensors placed in an office chair (Xu et al. 2012), cushions placed on a chair with an integrated sensor system , cameras measuring motion (Ferreira, Caraban, and Karapanos 2014), or motion sensors (Mateevitsi et al. 2014). Physical activity is measured with the number of steps individuals take (ƒ = 7/40, 17,5%) or via their heart rate (ƒ = 4/40, 10%). Several approaches are employed to measure the steps of individuals including activity trackers (Brombacher et al. 2019;Stamhuis et al. 2021), connected applications (Harjuniemi et al. 2020), accelerometers with self-developed algorithms (Ren et al. 2019a), or pedometers (Fortmann et al. 2013;Lim et al. 2010). We observe that some systems use indirect ways of data collection where the actions of the users are translated into data. These systems collect the data, which are used in the intervention, automatically. Examples of this are Stimulight (Brombacher et al. 2019) where the number of steps of users is collected in an automated and indirect way using an activity tracker. In a similar, automated, and indirect way, Health bar (Mateevitsi et al. 2014), measures if the user is present or absent at their desk with sensors placed around that desk. This indirect and automated way of collecting data is also used with other types of data where Khot et al. (Khot et al. 2017;Khot, Hjorth, and Mueller 2014) measure the heartbeat of individuals to create 3D-generated artifacts.
Next to personal data, we see a group of design interventions measuring the work environment (ƒ = 4/40, 10%) or the occupants of the building (ƒ = 3/40, 7,5%). The work environment of individuals is measured to create awareness about the space in which individuals are working and/or nudge them to work in a healthier environment. This is accomplished by placing a sensor in the environment to learn about the behaviour of individuals (e.g. stair taking vs elevator taking Rogers et al. 2010). Environmental factors that are measured include light (ƒ = 3/40, 7,5%), sound (ƒ = 3/40, 7,5%), air quality (ƒ = 2/40, 5%), as well other factors such as temperature and humidity (ƒ = 2/40, 5%). These systems combine low-cost sensors, and the combined insights of sensors, to learn about the work environment. They collect data in an automated and indirect way where the action of users (e.g. changing the work environment) are directly translated into data. Environmental or work-related data is also used to stimulate people to have an active work style. For example, Office Agents (Stamhuis et al. 2021) uses both personal (steps and productivity) and environmental data (light, sound, and air quality) to give insights into the work environment of individuals. On a similarly personal level, Apphia (Brombacher et al. 2020) collects data on the productivity level to motivate office workers to have an Table 1. Corpus of field study papers coded by the categories target audience, stages of change, and processes of change. 6 active work style. Off-the-shelve programmes (such as RescueTime) are used in these systems where this software is used to analyse the work pattern and data is sent to artifacts to provide feedback or awareness.

Output
Output is categorised into two sub-categories: (i) medium (light, shape-change or movement, sound or voice, display-based and computer control (visualised in Figure 3)), and (ii) sense (sight, touch, smell, hearing, and taste). Design interventions have several output modules to motivate individuals to have an active and healthy work style. Light (ƒ = 18/40, 45%) is a modality commonly used to convey feedback to a user in a nonintrusive way (not distracting individuals from their work). These artifacts are developed as ambient light displays (e.g. Brombacher et al. 2020;Mateevitsi et al. 2014;Ren et al. 2019a), or integrated into lamps which are used as a medium to express the feedback in both individual settings (Fortmann et al. 2013; Moradi and Wiberg 2017) and social settings (Pereira et al. 2016;Rogers et al. 2010). Examples of this can be seen in LightSit (Ren et al. 2019b) which uses a sensor mat embedded into an office chair and a lighting display to indicate if a person is sitting too long or has a bad posture. NEAT-lamp (Moradi and Wiberg 2017) uses a similar approach by using light as a reminder that the worker has been stationary during this time. Next to this, shape-changing artifacts and movement (ƒ = 8/ 40, 20%) are used as feedback modalities to create awareness and make individuals reflect on their work style. The shape-changing aspect is used to form 3D printed objects based on the heart rate of individuals (Khot et al. 2017;Khot, Hjorth, and Mueller 2014) or shape-changing furniture which adapts based on the activity pattern of individuals (Damen et al. 2020c;Ferreira, Caraban, and Karapanos 2014;Fujita et al. 2021). Examples of this can be seen in Ivy (Damen et al. 2020e) and TiltChair (Fujita et al. 2021) where the chairs change during the day to make people reflect on their sitting time.
Sound or voice (ƒ = 4/40, 10%) are output modalities seen in office robots to give feedback more directly. This is observed in the voice output between users and robots (Brandstetter, Liebman, and London 2015;Kanaoka and Mutlu 2015;Sabanovic, Reeder, and Kechavarzi 2014). For example, Kanaoka et al., (Kanaoka and Mutlu 2015) developed a social robot that uses dialogue and nonverbal engagement to enable office workers to talk about and reflect on their reasons for their lack of motivation for being physically active. With a similar approach, Fidgebot (Brandstetter, Liebman, and London 2015) is developed a social robot that drops verbal cues to enhance social interactions where colleagues can have micro-exercises. Display-based outputs (ƒ = 7/40, 17,5%) are being used in several ways. The outputs are used in displays that are embedded in artifacts to visualise feedback and enable users' interaction with the artifacts (Damen et al. 2020c;Kirkham et al. 2013;Stamhuis et al. 2021). Artifacts are also combined with a digital app, where the artifact itself is tangible, but the feedback is provided through an app .
When looking at the senses that these output trigger, we see that most interventions target sight (ƒ = 35/40, 87,5%) of people. This sense is frequently chosen due to its non-obtrusive way of communicating. With  Table 2. Corpus of 40 papers coded by the categories input, output, time, form, space relation, target behaviour and user intervention. 8 light and movement, artifacts can create awareness and trigger individuals without distracting the entire office environment. Secondly, we see touch (ƒ = 11/40, 27,5%) as a sense used in artifacts. Touch is used to interact with an artifact (Damen et al. 2020c), move the artifact (Probst et al. 2013a), or measure the presence of users (Rogers et al. 2010). Touch is also used to trigger users to be more active as can be seen in Ivy (Damen et al. 2020e) and TiltChair (Fujita et al. 2021). In lesser/no extent, smell (ƒ = 0/40, 0%), hearing (ƒ = 6/40, 15%) and taste (ƒ = 2/40, 5%) are used in artifacts. Hearing is mostly seen in robot-related interventions using voice/speaking, while taste is seen in the work of Khot Khot et al. 2017;Khot et al. 2015 by creating sports drinks and chocolate based on the heart rate of individuals. Despite emerging research into olfactory interaction Maggioni et al. 2020, currently, no intervention targets smell as an output modality.

Form factor
Several forms are used when developing Office wellbeing-related interventions namely: desk objects, wearables, furniture, robots, 3D objects, and gadgets, or interventions being embedded in their environment ( Figure 4). Interventions are designed as desk objects (ƒ = 12/40, 30%) where the artifact is placed on the desk of individuals. In this way, individuals can see the design (and the awareness and/or feedback it gives) when working. These desk objects have dimensions (usually sized around 15×15 cm) so they fit on the desk of individuals without overcrowding and disturbing the workspace (Brombacher et al. 2019;Mateevitsi et al. 2014;Ren et al. 2019a;Ren et al. 2019b). This makes sure that individuals can still accomplish their work tasks, while simultaneously creating awareness on their activity pattern. Next to desk objects, there are wearables (ƒ = 2/40, 5%) where individuals wear, attach, or carry the artifact on them. With this form, individuals can always bring the intervention with them. Idle Strips shirt (Harjuniemi et al. 2020) and Pediluna (Lim et al. 2010) both created a wearable combined with light to create social awareness and cues to decrease the sitting time of individuals and have an active work style. The wearable aspect of the artifact makes it possible to always have the artifact with you, informing both the wearer and their surroundings.
Several pieces of furniture are used to create a more active and healthier work style (ƒ = 7/40, 17,5%). Chairs (Damen et al. 2020e;Fujita et al. 2021;Min et al. 2015;Probst et al. 2013b;Xu et al. 2012) are chosen mainly as a medium to learn about the posture and sitting time of individuals. Next to chairs, we see a group of customdeveloped pieces of furniture, such as PositionPeak (Damen et al. 2020b) and the shape-changing table of Grønbaek (Grønbaek et al. 2017), which triggers individuals to have more active work styles and meetings (Damen et al. 2020a). Robots (ƒ = 4/40, 10%) are also used as a social medium to trigger active workstyles. These robots (Brandstetter, Liebman, and London 2015;Kanaoka and Mutlu 2015;Sabanovic, Reeder, and Kechavarzi 2014) have the ability to physically roam around (creating more attention than stationary objects) and have a more dynamic role (creating social interaction between human-robot and human-human) due to them not being bound to a person or environment. Finally, we identify a group of designs that are embedded in the environment (ƒ = 6/40, 15%). The embeddedness of these interventions is exemplified in them being (fully) integrated into meetings ( (Mateevitsi et al. 2014), shape-changing/movement (Reeder et al. 2010), voice/sound (Sabanovic, Reeder, and Kechavarzi 2014), digital outputs (Stamhuis et al. 2021), and computer control (Probst et al. 2013b).
Kay 2020), buildings (Rogers et al. 2010), or work settings (Damen et al. 2021). These interventions trigger a larger audience, due to them being visible in an open environment (Damen et al. 2020d;Damen et al. 2021) or assisting with active group meetings (Damen et al. 2020c). This can be to evoke active ways of working like WorkWalk (Damen et al. 2020d) or to make individuals reflect on using the elevator (Rogers et al. 2010).

Space relation
The design interventions presented in this paper are, because of the focus on the domain, all situated in the office space ( Figure 5). Most of these interventions are introduced as new products in the office environment (ƒ = 33/40, 82,5%) and are placed directly in the work environment, desk, or chair of individuals. Additionally, some artifacts are added to existing office products as 'Additional Object' (ƒ = 2/40, 5%) or which replace existing products (ƒ = 5/40, 12,5%). The replacements are often new and active pieces of furniture that are complementary to the existing non-interactive furniture (Damen et al. 2020b;Xu et al. 2012).
Interventions developed to improve the physical activity of office workers have spatial outputs and visibility. On an ego-centric level, approaches focus on individuals themselves. As exemplified in interventions such as Pretty Pelvis, sensors integrated into the users' seats give hidden, personal feedback on the posture via an avatar in an app . On an individual, place centric-level, artifacts such as the Healthbar (Mateevitsi et al. 2014) and the Idle stripes shirt (Harjuniemi et al. 2020) give feedback on the physical activity level. These interventions however visualise their data in an open environment, making it visible for others to see and comment on. On a social, place-centric level, Stepby-Step (Ren et al. 2019a) is a gift-like artifact that office workers share to challenge each other to improve physical movements and social interactions. Stimulight (Brombacher et al. 2019) takes a different social approach by visualising social feedback about the physical activity of individuals in a non-competitive and non-milestone way, triggering also individuals in the work environment that are not motivated by competition. These artifacts visualise their data (often via a light or shape-changing mechanisms) in the periphery of individuals switching between the centre and peripheral attention of users.
We see a group of these interventions being bound to their location, either being bound to the desk (ƒ = 8/40, 20%) or bound to the environment (ƒ = 14/40, 35%). The desk-bound artifacts are bound to the desk due to their size (Stamhuis et al. 2021), power management (Brombacher et al. 2020), or sensor being attached to the desk (Mateevitsi et al. 2014;Ren et al. 2019b). This can be seen in Lightsit (Ren et al. 2019b) where the feedback mechanism of the artifact is implemented in the screen stand, giving it a desk-bound structure. These interventions fall in the action space of individuals where they are accessible for the user to interact with the artifact. Environment-bound artifacts are specifically designed for a certain space like the coffee table (Kirkham et al. 2013), stairway (Rogers et al. 2010), or meeting room (Damen et al. 2020b). An example of this can be seen in the work of Rogers (Rogers et al. 2010), where an intervention is bound to its environment due to the purpose of the design (making individuals reflect on their elevator/staircase use).
Next to the location-bound artifacts, we see a more dynamic group of artifacts that can be moved or move themselves. Chairs and cushions, which can be placed on chairs (ƒ = 4/40, 10%) are used as objects where individuals can replace them themselves depending on their working location (e.g. Fujita et al. 2021;Min et al. 2015). Some of the robot-like interventions (ƒ = 2/40, 5%) move themselves throughout the office, creating a more dynamic situation by involving/triggering multiple people and locations. We also see a group of 'Bring along items' (ƒ = 9/40, 22,5%), which are smaller items and/or wearables. These interventions have the option that individuals can easily bring them with them to a different location within the work environment (or even at home). The bring-along items are mostly compact interventions in which all components (power source, sensors, and output module) are included in the intervention (Brombacher et al. 2019;Ludden and Meekhof 2016;Moradi and Wiberg 2017;Ren et al. 2019a;Züger et al. 2017). A similar approach is seen in wearable artifacts (Harjuniemi et al. 2020;Lim et al. 2010) where compact designs are placed on individuals.

(Exposure) time
Office well-being interventions take different approaches when it comes to time. With time we indicate when individuals have access or are exposed to the artifact. Most of the developed interventions focus on the working time of individuals (e.g. between 9-5, ƒ = 29/40, 72,5%). These interventions are placed in the office and can't be moved to other non-working locations. A smaller group of artifacts takes the complete day (24/7, ƒ = 6/40, 15%) into account. These artifacts collect data from the user throughout the day and visualise this on the design directly like Idle Stripes (Harjuniemi et al. 2020) and Pediluna (Lim et al. 2010), or collect the data and use this to form the artifact (Khot et al. 2017;Khot, Hjorth, and Mueller 2014).
At last, we see a category of interventions specifically focusing on meetings (ƒ = 5/40, 12,5%). These artifacts are triggering/facilitating for active meetings (Damen et al. 2020b;Damen et al. 2020d) or collect data about the meeting environment and use this to increase productivity (Constantinides et al. 2020).
Several strategies are used in the frequency of the data presentation or visualisation within the intervention. Interventions use certain thresholds, such as Stars (decibels in the work environment Pereira et al. 2016) and Apphia (productive minutes Brombacher et al. 2020), after which the data is presented (Brombacher et al. 2020;Harjuniemi et al. 2020;Mateevitsi et al. 2014;Pereira et al. 2016;Reeder et al. 2010). Others collect data in a set time interval after which the data is presented directly (Brombacher et al. 2019;Ferreira, Caraban, and Karapanos 2014;Fortmann et al. 2013). Examples of this are where feedback is provided after a certain number of inactive minutes (Mateevitsi et al. 2014) or presenting feedback every hour on the activity pattern of individuals (Brombacher et al. 2019). We also identified artifacts that sense continuously throughout the day and present their data once at the end of the day (Güldenpfennig, Ganhör, and Fitzpatrick 2015;Khot et al. 2015;Khot et al. 2017;Khot, Hjorth, and Mueller 2014), or that present their feedback continuously throughout the day (Lim et al. 2010;Züger et al. 2017).

User interaction
When analysing the user interaction ( Figure 6) with Office well-being intervention, we see a large section of the artifacts simply not having any form of user interaction (ƒ = 18/40, 45%). Most of the artifacts collect data, either from the user or its environment, and have an output linked to this data. The collection of data is an automated and indirect way without any input or interference from the user. The user cannot interact or negotiate with the artifact and therefore does not have direct, user-controlled input. The interventions that have a form of user interaction see this translated in interactions such as: voice/speaking, changing the position of design and/or interacting with it, interacting with the environment, taking steps with the artifact, or changing posture ( Figure 5). The artifacts allow user interaction, often having a form of interaction, where individuals change the position of the artifacts and/or press the artifacts (ƒ = 10/40, 10%). This interaction is implemented in digital interface interaction (Damen et al. 2020c;Khot et al. 2015), snoozing (Reeder et al. 2010), providing feedback (Gallacher et al. 2015;Karlesky and Isbister 2013), or changing the position of the artifact (Ludden and Meekhof 2016;Stamhuis et al. 2021). The interventions have a direct and user-controlled interaction in the action space of individuals. Mood Squeezer (Gallacher et al. 2015) and Fidget Widgets (Karlesky and Isbister 2013) use squeezing and fidgeting for users to interact with their artifacts. Next to this, we see designs such as Office Agents (Stamhuis et al. 2021) and Pebble (Ludden and Meekhof 2016) where users move their artifact to improve their working environment. Additionally, we identify artifacts improving the physical activity pattern of individuals by taking steps with the artifact (ƒ = 2/40, 5%) or change the posture of individuals (ƒ = 4/40, 10%). The posture change interaction is linked with chairs or cushions which measure posture change (Fujita et al. 2021;Min et al. 2015;Ren et al. 2019b;Xu et al. 2012). These forms of user interaction have an indirect approach where the action or activity of the user is translated into data.
While most of the artifacts have direct user interaction. We also see a group of artifacts where the user interacts with the environment (ƒ = 3/40, 7,5%). Interventions such as The Clouds and Follow-the-Lights (Rogers et al. 2010) have sensors implemented in the environment to learn about the behaviour of individuals and use this to nudge them to have an active work style.

The stage of behavior change
The studies were analysed based on the duration of the study and behaviour change techniques that are used in the intervention (based on the 10 processes of change, Table 1). After placing the artifacts within the Transtheoretical Model (Prochaska, Rodgers, and Sallis 2002), we see a large group of artifacts in between the preparation/action stage of change (ƒ = 10/24, 41,7%). Most of these interventions are tested with a small group of individuals over a short time (1-8 weeks). These interventions are evaluated to learn about the first reaction of the user and some initial hints of behaviour change are seen. Counterconditioning is one of the most common strategies being used by substituting healthy alternative ways in the artifact (Constantinides et al. 2020;Damen et al. 2020b;Damen et al. 2020c;Güldenpfennig, Ganhör, and Fitzpatrick 2015;Rogers et al. 2010;Züger et al. 2017). This process of change is combined with stimulus control to further enhance the behaviour change process (Damen et al. 2020b;Damen et al. 2020d;Rogers et al. 2010).
Secondly, we see a group of interventions that fall in between the contemplation and preparation stage of change (ƒ = 5/24, 20,8%). Individuals are intending to start a healthier behaviour and had some initial experience with the artifact. No intention of behaviour change is however observed in the studies. Consciousness-raising is seen as a common strategy in behaviour change techniques where the artifact plays a role in making individuals aware of their unhealthy work habits (Fortmann et al. 2013;Fujita et al. 2021;Mateevitsi et al. 2014;Sabanovic, Reeder, and Kechavarzi 2014). Getting help from others (social liberation) is a final strategy seen in these interventions where the social work environment is used as a start to intent a healthy behaviour. Figure 7  (Züger et al. 2017). WorkWalk is the only study that is going into the maintenance stage where individuals have changed their way of working and regularly take 'Workwalks' to replace their regularly sitting-focused meetings. Figure 8.
Many design interventions focus on creating a healthier and more active work style. This is often done by triggering users to take breaks (ƒ = 8/40, 20%) or motivating them to be physically active (ƒ = 9/40 22,5%). This increase in physical activity is measured in steps (Brombacher et al. 2019;Fortmann et al. 2013;Lim et al. 2010), doing micro-exercises (Brandstetter, Liebman, and London 2015), or moving more actively throughout the office building (Rogers et al. 2010). Next to this, we see a group of designs that do not directly want to trigger more active ways of working, but that want to create awareness about the inactive behaviour of individuals (ƒ = 12/40, 30%) or want them to reflect on it (ƒ = 4/40, 10%). This awareness is principally created via light (Brombacher et    about the workstyle of individuals in a non-intrusive way, creating the possibility for participants to reflect on personal feedback. The current field of interventions focuses on breaktaking, instead of creating more active ways of working (ƒ = 6/40, 15%). The examples that do take this approach are seen in PositionPeak (Damen et al. 2020b) and the interactive office chair (Probst et al. 2013b) where active pieces of furniture are designed to promote an active work style. Another example, Work-Walk (Damen et al. 2020d) was developed as a 25-min route to encourage walking meetings. These interventions highlight ways to integrate active ways of working, without actively taking a break from work activities.

Summary of findings
Underneath, we summarised the key findings from the analysis of the Tangible Office well-being interventions: Limited use of multiple data sources: most design interventions make use of a single data source which are steps or user presence and movement. The dominant focus is on collecting personal data with limited work exploring the building or the entire work environment. Several approaches are seen in both the use of individual, social, and combination of data, triggering several interpersonal aspects.
Similar approaches in output mediums: a similar approach is seen in Office well-being interventions using light (and to a lesser extent shape changing/movement) as a data visualisation medium. Artifacts trigger the periphery of the user focusing on the sight and/or touch of individuals. These mediums are chosen to create awareness (both to individuals and their colleagues/ environment) without distracting them directly from their work.
Design outputs in interventions: Office well-being interventions are often developed as desk objects where the artifacts are placed in the direct environment and proximity of individuals. Next to this, there is a group of interventions that integrate their artifact within the environment by developing pieces of furniture or which move around in the environment in the form of robots.
Artifacts being bound to their environment: most office design interventions are introduced as new products that are bound to the office environment. These artifacts lack the option to be moved to other office environments or the option to bring them with you throughout the day, including the home environment. The artifacts that are not bound to their environment, are often seen in chairs or robots which can be replaced (or move themselves) in the office.
Lack of user interaction: there is a lack of user interaction in Office well-being interaction. This lack of interaction removes the option to negotiate or provide feedback to the system. The artifacts have a form of interaction based on the movement and/or pressing to the artifact or have an active work style with the artifact by taking steps with it or changing posture.
Lack of behaviour change foundation: The goal of the majority of Office well-being interventions is to instigate healthier and active ways of working. Interventions are mostly evaluated over a short period with a small group of individuals, therefore only reaching the early stages of behaviour change. This is however mostly based on behaviour change techniques (e.g. awareness or social support), but there is a lack of grounding based on theoretical models (Orji and Moffatt 2018).

Design considerations
The 40 design interventions were analysed based on the 7 classification areas. We combine findings and relate these to theories (e.g. periphery attention, job crafting, social feedback, or user interaction) and future trends (e.g. hybrid working environments) in office environments. Based on the combination, we give design considerations for the development of future Office wellbeing design interventions.

Design consideration 1: what data to capture?
We see three types of data that are collected within Office well-being design interventions, namely personal data (e.g. steps, sitting time, posture, and heart rate), work data (e.g. productivity), and environmental data (e.g. light, sound, and air quality) on both individual and social levels. These designs however often rely on a single type of data (with some exceptions such as Office Agents Stamhuis et al. 2021). There is however a growing group of work, researching the relationship between these three types of data such as step count, temperature and heart rate data (van Kasteren, Champion, and Perimal-Lewis 2019), environmental (noise, air quality, light intensity, etc.), physiological variables (heart rate) and breaks (Punait and Lewis 2019) or combining computer interaction, heart-, sleep-, and physical activity-related data (Züger et al. 2018). These examples do however miss the tangible design component in their work. Combining several forms of data within a tangible design could therefore have a bigger effect when improving the productivity and well-being of office workers (van der Valk et al. 2015).
To answer the question 'What data to capture?', we see opportunities in combining personal, work, and environmental-related data to learn how these can create healthier and active work styles. This will also gain insights into possible trade-offs between data types such as physical activity and productivity. Where individuals take breaks from work to increase their productivity over the day, while also having the tradeoff that they take a break from work, which will make people less productive. With the collection also comes the Human-data interaction and giving personal value to the collection of data (Mortier et al. 2014). Currently, data is to a big extent collected passively without any control of the user. To give insight and value to the collection of data, individuals should have the option to have a clear understanding of how the data is collected and used, have control over the use of the data and the societal contract surrounding the use of data (Mortier et al. 2014). This will give the user a more central role in the collection of data, giving more value to understand to both users and researchers.

Design consideration 2: form factor (spatial application)
A majority of the Office well-being interventions are bound to the office environment itself. We see this in the desk and environmental-bound artifacts, as well as the time that these artifacts are active, which is only during office hours. Society is however seeing a shift from working in offices to a hybrid setting of both working from home and office after the COVID-19 pandemic (Jiang et al. 2021;Wang et al. 2021). Research indicates a preference of employees who prefer to see this trend continuing post-Covid (Bloom et al. 2020). This change to a hybrid form of working (office and home) asks for hybrid artifacts. These interventions should be able to collect data and present their output throughout the changing environment of users. These hybrid artifacts and approaches are mostly seen in applications and wearables (Jiang et al. 2021), but we would motivate the research community to explore how these hybrid artifacts can be developed as tangible, technological interventions which help to create a healthy and active work style. These tangible systems can be developed as the earlier mentioned, but further developed, 'bring along items', wearables, or hybrid systems where users have devices at several locations where they can check-in. This is, however, an under-explored field and we can currently only speculate on the effect of these future artifacts.
With this comes the temporality of the artifacts. Most of the interventions function during the office hours of individuals (e.g. 9-5) and therefore do not consider the activity pattern of individuals during the whole day. Examples of this can be seen when individuals cycle to work and directly after arriving at the office, the artifact gives feedback that the user should be active. Future artifacts should therefore be tailored to be contextaware and use multiple sources of data so that the correct feedback is delivered at an opportune moment (Huang, Benford, and Blake 2019;Nahum-Shani et al. 2018). The field of machine learning and artificial intelligence can play an important role where artifacts can adapt and learn based on the habits and agendas of individuals (Rabbi et al. 2015).

Design consideration 3: user interaction for richer experiences
Interactions between users and Office well-being design interventions are often missing. The artifact collects data about the users and/or their environment via embedded sensors and feeds this back to the user by visualising the data to provide insights into their work style. These 'Collectors' (Cila et al. 2017) have the potential to change the behaviour of the users but cannot measure this. The user also cannot interact with the artifact, removing the option to provide feedback, negotiate or overrule the system (Rozendaal 2016). Failing to engage with the product could therefore lead to growing tension between the interventions and the users (Cila et al. 2017). The input of users is also needed in the future development of machine learning in these systems where the 'human in the loop' is needed to direct the processes of the system (Zanzotto 2019).
Adding a direct, user-controlled interaction within the action space of individuals could also enrich the overall experience with the artifact. If users can give feedback when they, for example, work in their ideal work environment, they can indicate this within the artifact. These parameters can then be collected and based in a larger database where machine learning can be used to recommend healthier and personalised work settings, depending on tasks, work demands, or preferences (Rabbi et al. 2015).
The future development of Office well-being design interventions should focus on smart products which act autonomously with the behaviour of users. These artifacts should be able to learn from the users in both direct (user feedback/interaction) and indirect (collecting data from the users and their environment) and combine them to provide insights and measure potential behaviour change of users. These systems should not only focus on the artifact itself but also the platforms, infrastructures, and technologies around the artifact, creating a coherent experience (Knutsen 2014).

Design consideration 4: designing for active ways of working
The goal of Office well-being design interventions is to create a healthier and more active work style. Interventions take the approach of 'Taking breaks' and by doing so, increasing the physical activity level, productivity or breaking the sedentary work style of individuals. These artifacts, however, don't change the way individuals work, but merely support taking a break from unhealthy work patterns, instead of facilitating more active ways of working (Damen et al. 2020a). Only a couple of interventions facilitate active ways of working, mostly seen in the work of Damen (Damen et al. 2020b;Damen et al. 2020c;Damen et al. 2020d) and Probst (Probst et al. 2013a;Probst et al. 2013b). These interventions however often miss a data input to learn about the behaviour of users. The lack of data input makes it hard to learn about the behaviour of users, especially on a longitudinal level where you cannot constantly observe the user. We, therefore, recommend the research community to expand the work to design for active ways of working by taking a data-driven approach to learn about the behaviour of users.

Design consideration 5: job crafting and employers' attitude towards active work environments
The analysed Office well-being interventions are introduced as a finished artifact to a certain group of individuals in the office environment. These interventions may however not directly fit (or even conflict) with the work pattern of individuals. The work environment is changing where job crafting is playing a central role. Job crafting is defined as 'an employee-initiated approach that enables employees to shape their own work environment such that it fits their individual needs by adjusting the prevailing job demands and resources' (Tims and Bakker 2010). The enabling of employees' own work pattern, should also include being able to have a physically active work style. Job crafting plays here an important role in future ways of working where time-spatial flexibility is used to create a person-job fit where individuals can plan their own healthy and active work patterns (Wessels et al. 2019).
Before active ways of working could be effectively implemented, an initiative is needed to set up the foundations of change at the organisational level (Parks and Steelman 2008;Renner et al. 2020;Sandblad et al. 2003;Wang, Jiang, and Chern 2014). This should also help with removing the stigma associated with 'the workplace is only for working', thus creating favourable conditions for a higher acceptance of being active on the work floor (Lin et al. 2006).

Design consideration 6: periphery attention in output and sense
A majority of the interventions focus on the peripheral interaction between the artifact and the user (Bakker, van den Hoven, and Eggen 2015). The artifact switches between the centre and periphery attention of users by providing feedback/awareness. The switch between attention mediums gives the benefit of nudging users while being able to perform other activities in parallel (for example: working behind your desk while receiving feedback on your posture Reeder et al. 2010). Light and shape change/movement are therefore often used as medium, due to them triggering the sight of individuals with this easy shift between the periphery and centre of attention. These systems have a place-centric-approach, which enables visibility in an open office environment, creating the possibility to give social feedback on the output of the intervention (Brombacher et al. 2019) The output from these artifacts mostly use a onesided approach. This approach, as described above, is chosen due to it creating awareness without disturbing the whole work environment. Output mediums such as taste and smell are less frequent (or not used) due to them possibly distracting others. These mediums can however be used as information decoration, where individuals to whom the information is not relevant, can still benefit from the presentation/visualisation of data (Eggen and Van Mensvoort 2009). In this way, the feedback is not intrusive for the whole environment yet provides information for the involved users. This is especially seen in the emerging field of olfactory-focused design which shows promising areas where smell is used as an interaction modality in design (Maggioni et al. 2020).

Design consideration 7: social, individual, or both and what about privacy?
Interventions have different scopes, focusing on individuals, groups, or both. Several strategies are seen in both individual and group/socially focused artifacts. On an individual level, interventions can present their feedback purely to the user, hidden from the environment. Next to this, we see a group of individually focused artifacts that present their feedback in an open environment. The visibility in the open environment makes it possible for others to comment, creating the possibility to also give social feedback (Brombacher et al. 2019) Interventions use social support as it has been consistently linked and indicated as an important factor to improve the physical activity level of individuals (Edwards et al. 2014;McNeill et al. 2006;Meske and Junglas 2021;Orji and Moffatt 2018;Prochaska, Johnson, and Lee 2009;Vaziri et al. 2020). Several strategies can be used to evoke social support such as milestones, cooperation, and competition (Brombacher et al. 2019). The office environment is, however, a diverse environment of people with different levels of competitiveness. Implementing a competition and/or milestone-focused artifact could therefore only trigger competitive-minded people (Fletcher, Major, and Davis 2008). Designers should therefore understand the background and personality of office workers, before implementing certain social support techniques.
When feedback is presented in an open environment, privacy concerns emerge. Data related to physical activity, well-being, productivity, or the work environment are suddenly 'visible' or accessible for others to see and comment on. Ethical considerations should be considered so that this information is not misused by both employers and colleagues (e.g. controlling if people are working hard enough) (Edwards et al. 2014). Transparency is therefore needed on who has access to the data and where the data is stored (i.e. Human-data interaction) (Edwards et al. 2014;Mortier et al. 2014). Data physicalization can also play a role there where data is visualised aesthetically, only giving meaning to the user and the people they chose to tell (Moere and Hoinkis 2006).

Design consideration 8: designing for behavior change
Most Office well-being interventions have the goal create healthier and active work style for office workers. Most interventions, however, currently do not reach this level of behaviour change, with most design interventions only being evaluated over a short period. The artifacts often lack an explicit foundation in theoretical models and behaviour change techniques (Damen et al. 2020a;Huang, Benford, and Blake 2019;Orji and Moffatt 2018). Comparing them to the Transtheoretical Model (Prochaska, Johnson, and Lee 2009) also shows that interventions are almost all being placed in the initial stages and almost non reaching the maintenance stage.
Ethical implications are an ongoing debate when using theocratical grounding to initiate potential behaviour change. While this is a general concern in behaviour change interventions, it should also be considered in tangible Office well-being interventions. Not respecting the right of autonomy of individuals could risk reducing their ability for autonomy, and risks increasing health inequalities (Tengland 2012). To avoid ethical problems and to increase autonomy is to respect the individuals' right to self-determination and make sure they are involved in the process including the problem formulation, the decision process, and actions (Bijleveld, Andries, and Van Rijckevorsel 2000;Meske and Junglas 2021;Rapp 2019;Tengland 2012).
Office well-being interventions need to have a different approach here when behaviour change is the goal. First, a grounding, in theory, needs to be taken when developing the interventions. Based on the chosen theory, behaviour change techniques should be chosen to facilitate the behaviour change towards healthier and active work styles (Munir et al. 2018;Orji and Moffatt 2018). These interventions should be evaluated over a longitudinal period to learn if and how the behaviours of individuals changed.
Evaluating artifacts over a longitudinal period asks for interventions that have multiple user-system interactions based on the behaviour of the user toward the design (Lockton et al.). These interactions should be based on persuasion principles (Oinas-Kukkonen and Harjumaa 2009) where several design principles are implemented (or combined) in the design. The artifact field also sees an opportunity here to research the effect of using multiple design principles to learn how these can affect the behaviour of individuals (Michie, van Stralen, and West 2011;Orji and Moffatt 2018).

Discussion
The goal of our work is to define the field of tangible Office well-being interventions using 7 classifications that were informed by literature on the topic. Our work is a starting point for further research into the topic of Office well-being. In the discussion, we reflect on (i) the role of tangible user interfaces in Office well-being interventions, (ii) the goal and use of behaviour change techniques, (iii) relation to work practices, and (iv) the use and future development of the classifications.

Role of tangible user interfaces in office wellbeing interventions
At the start of this meta-analysis, we identified a fragmented research landscape on tangible Office wellbeing interventions, missing the relationship between interventions, data, design strategies, and outcomes and behaviour change techniques. This led to the question: how can tangible office design interventions play a role in the learning and measuring (of individuals and their environment), understanding (response towards the design), and role of data visualisations or presentation in creating healthier and active ways of working? Reflecting on this question, we see a tangible approach in the interventions, which differs from commonly used apps or activity tracker devices, that extend the possibilities to display and visualise data. The tangibility of the artifacts creates awareness in the periphery of individuals (Bakker, van den Hoven, and Eggen 2015) and the possibility for social support between individuals (Brombacher et al. 2019), while, also being part of the physical office environment (Shaer and Hornecker 2009). When analysing the properties of Tangible Office well-being, we see 3 missing properties in current interventions: robustness and reliability, scalability, and portability (Shaer and Hornecker 2009). These properties need to be addressed to enable new hybrid ways of working (portability) (Wang et al. 2021) and conduct longitudinal studies (scalability, robustness, and reliability) in the future development and studies of Office well-being interventions.
The learning and measuring from the user and their environments are currently mostly done via a single data source either being personal, work, or environmental-related data. Data is collected with both offthe-shelf sensors/software (e.g. activity trackers or Res-cueTime) and in some cases self-developed sensor systems. These sensor-controlled systems often lack the possibility for user interaction leading to a lack of control (Rogers and Muller 2006). A key data source that is currently missing is the direct user input/interaction with the artifact. This leads to a lack of understanding of the use and interaction of the intervention, resulting in a shortage of empirical data, collected with the artifact. It also creates a communication issue where the 'invisible' data stream is collected without any opportunity for the user to react and interpret the data stream. (Bellotti et al. 2002). The field of tangible interaction provides an opportunity here where systems rely on embodied interaction, tangible manipulation, physical representation of data, and embeddedness in real space to an interaction between users and Office wellbeing interventions (Hornecker and Buur 2006).
A similar approach is observed in the medium of visualisation and/or presentation where light is chosen to present the data. The use of light however provides minimal physical feedback in tangible user interfaces (Shaer and Hornecker 2009). The emerging field of data physicalization can play a key role here where it can bridge the field of visualisation and tangible user interfaces (Jansen et al. 2015). This field sees opportunities for developing shape-changing physicalizations for tangible and human-computer interaction. This field of Office well-being interventions is also moving towards a new promising direction with machine learning and artificial intelligence being used to classify activities and generate suggestions for healthier and active work styles (Rabbi et al. 2015). These interventions will combine multiple data sources and user interaction in creating context-aware intervention which should deliver their feedback at an opportune moment when individuals have the opportunity to be physically active (Chung et al. 2017).

Goal and use of behavior change theories in office well-being interventions
While this paper does not specifically, from the literature perspective, explores behaviour change theories, both previous analyses (Damen et al. 2020a;Huang, Benford, and Blake 2019;Orji and Moffatt 2018) and our work show a lack of theoretical foundation in Office wellbeing design interventions. Designers must identify for which behaviour change phase they are designing (Prochaska, Johnson, and Lee 2009) and the overall goal of the research. Before developing an artifact, an understanding of the behaviour in context is needed. This includes an understanding of the individual, social and physical environment before a behavioural target can be achieved (Michie, van Stralen, and West 2011). Based on the understanding of the context and the behaviour, functions of the design intervention can be developed and implemented which can help to change the selected target behaviour (Michie, van Stralen, and West 2011;Munir et al. 2018;Orji and Moffatt 2018). The developed artifact should be evaluated over a longitudinal period to learn if and how the behaviour of individuals changed, something which is currently missing in the evaluation of most Office well-being interventions.
Behaviour change is, however, not always the goal of Office well-being design interventions. Several designs are developed as awareness tools to facilitate exploration and learn if a certain interaction, visualisation, or technology has an initial effect on the work pattern of individuals. A differentiation should therefore be made about the overall goal of the artifact. These explorative studies also often have a different approach with shorter user evaluation with a small sample size (Caine 2016). These small sample sizes are often enough to reveal basic usability problems (Lance and Vandenberg 2009) and need a smaller initial investment (Bacchetti, Deeks, and McCune 2011), but are not able to demonstrate or validate bigger goals or claims.
The current research field of Office well-being design interventions focuses for a large part on short and explorative studies. These studies provide a widespread exploration of different output mediums, data collections and visualisations, target behaviour, user interactions, and more. Our field is, however, ready for the next step, taking a new longitudinal approach, evaluating interventions over a longer period focusing on the potential behaviour change toward healthier and active work styles. This longitudinal approach asks for a different strategy where artifacts need to be capable to sustain longer periods of user interaction (e.g. firmness, data collection and storage, power management) as well as a grounding, in theory, to guide individuals throughout different stages of behaviour change (Prochaska, Johnson, and Lee 2009).

Relation to work practices
The office environment is changing at a rapid pace, and this is even more accelerated due to the Covid-pandemic. This change has started with individuals adopting a hybrid work setting where they are both working from home and in the office (Wang et al. 2021). Future offices see a continuing pattern here where digital advancements will combine working from home with meeting physically to sustain the community aspect to meet in real life (Newbold et al. 2021). This could also reduce the need for fixed working spaces and create a need for flexible individualised working environments. This individualised working pattern is already more common where companies have implemented flexible working hours, where individuals can schedule their work hours based on their work-life balance (Chung et al. 2017;Newbold et al. 2021). This fast and continuously changing work environment asks for new approaches and strategies in the development of Office well-being interventions.
The organisation/company plays a key role in the implementation of Office well-being design interventions. Before an artifact could be effectively implemented, an initiative is needed to set up the foundations of change at the organisational level (Munir et al. 2018;Parks and Steelman 2008;Sandblad et al. 2003). This positive organisational attitude towards an active workstyle will help with removing the stigma of 'the workplace is only for working' creating favourable conditions for a higher acceptance of being active on the work floor (Lin et al. 2006).

Use and future development of office wellbeing classification
Our current analysis provides an initial overview based on the classifications. This overview could, however, have a deeper layer of analysis. Such a deeper, more detailed, layer of analysis can be given for some of the (sub) classifications. For example, the medium of how steps as a data input are collected (e.g. accelerometer Ren et al. 2019a;activity tracker Brombacher et al. 2019;pedometer Fortmann et al. 2013;app Harjuniemi et al. 2020). Similarly, the space relation classification could be further analysed in relation to models such as the Proxemic Interactions (Greenberg et al. 2011) or Situative Space model (Surie et al. 2010) or from a user interaction level, the agency of the intervention towards data collection and the relations towards other interventions/products and users (Cila et al. 2017).
Based on the classification, 7 design considerations were formed. The considerations are set out as starting point for the future development of Office well-being design interventions. These considerations should however not be seen as the golden ticket where following all considerations will lead to the perfect design intervention. Considerations could contradict each other when implementing them in new artifacts. For example, when developing a tangible, open and socially focused design visualisation for periphery attention (consideration 6), while hiding information for privacy reasons (consideration 7). Future research with the implementation of the given considerations should give a better perspective on the impact and use of the considerations on Office well-being design interventions within the HCI community. But for now, these considerations are carefully positioned findings from literature that form various perspectives on tangible Office well-being.

Limitations
Our research consists of 40 selected Office well-being design interventions that are published in academic literature. This field does, however, not solely focus on academic literature but also on commercial products. Industry has also addressed some of the subproblems of Office well-being, focusing on individual aspects, but as a research field, we are more interested and complementary with our holistic view.
We evaluated the design intervention concerning their behaviour level state of change based on the Transtheoretical model (Prochaska, Johnson, and Lee 2009). This model and the stages of change could be a 'black box' where the transition toward the next stage is discussed, rather than understanding how the individual experiences each stage (Rapp 2019). However, this model was chosen due to it being the most used theoretical model in Office well-being interventions (Damen et al. 2020a;Huang, Benford, and Blake 2019;Orji and Moffatt 2018). This does have the consequence that the interventions were placed in the classification based on the coding and analysis of the authors. This was needed due to the lack of documentation on behaviour change techniques and the overall evaluation of the artifact concerning behaviour change. The mapping of intervention is therefore mostly developed as a first step in mapping Office well-being interventions and findings a common pattern in used behaviour changes strategies.
The Office well-being design interventions are classified into 7 categories. However, there are near infinitive ways to classify interventions (Michie, van Stralen, and West 2011) meaning there will therefore be different ways to classify these interventions and we would encourage the research community to further explore this topic. There is currently no overacting classification for this field. With our research, we contribute to the field of Behaviour & Information Technology (BiT), by providing design considerations for both the development of design interventions and research setups (Bijleveld, Andries, and Van Rijckevorsel 2000;Edwards et al. 2014). We hope that this work presents a starting for future research into tangible Office wellbeing interventions to design for a healthier and active work environment.

Conclusion
Office well-being has become an emerging topic where technology and design are combined to support and mediate healthy and active office environments. Our research identified a fragmented research landscape on tangible Office well-being design interventions, missing the relationship between interventions, data, design strategies, and outcomes and behaviour change techniques. This paper defines the field of Office well-being interventions by 7 design classification areas and their theoretical foundation in behaviour change. Based on this classification, we provide an overview, reflect on the opportunities and challenges, and give design considerations for the development of future tangible Office well-being interventions. Researchers can use this work as a starting point for future research into tangible Office well-being interventions to design for a healthier and active work environment.