Much of the academic literature on flow is based upon the work by Csikszentmihalyi. In his first publication on the matter (1975/2000), he identified the two main conditions needed for flow to occur:
Firstly, flow occurs when your skills (your capacity for action) and the challenges (opportunities for action) are in perfect balance. This balance fragile in nature:
Another contributor to this fragile balance is the fact that using your skills will improve those skills, decrease the perceived challenge, and thus reduce flow. In order to continue experiencing flow, more and more difficult challenges must be sought out to match the new skills. A good flow activity provides challenges that stretch existing skills and a system of progression (Nakamura and Csikszentmihalyi, 2009).
Notably, flow experience is determined not by the objective skills a person has or the challenges an environment poses, but rather the subjective perception of those skills and challenges. Because of this, while some flow activities provide goals and structures that make flow more likely, any activity can be flow-inducing for the right person.
Secondly, flow requires clear goals and immediate feedback on the progress made towards the next goals. In many flow activities, such as rock climbing or improvised music creation, the next step in the process is decided on-the-fly, based on previous steps: the next rock to grab onto, or the next note to play, arises from the interaction itself — a natural "next stepping stone".
Csikszentmihalyi (2009) calls these stepping stones proximal goals. If someone perceives, in the moment, something to be possible and worthwhile of exploration, that is set as a new goal and, in doing so, creates new proximal goals.
Proximal goals cause emergent motivation: instead of a predetermined script, the activity itself will motivate the user to perform the activity in a certain way. The direction of the activity is decided through a conversation between the person and the environment they're interacting with. In some cases this emergent motivation is felt as a literal conversation, such as between a painter and their work.
In order for flow to occur, there needs to be a constant ongoing interaction between the user and the environment. When the user acts, the environment needs to provide immediate feedback on the result of that action, so that the user can set the next goal.
While flow is observed across culture, class, gender, age, and types of activity (Nakamura and Csikszentmihalyi, 2009), not every person experiences the same amount of flow.
The capacity to experience flow appears to be nearly universal. Nevertheless, people vary widely in the frequency of reported flow.
Different people experience more or less enjoyment out of an activity, even if their skills and challenges are the same (Adlai-Gail, cited in Emerson, 1998). Csikszentmihalyi (cited in Emerson, 1998) calls this the autotelic personality – similar to how people have varying levels of dispositional positive affectivity (Meehl, 1975). The autotelic personality is characterised by a general curiosity and interest in life, persistence, and low self-centredness. What's more, these individuals seem more comfortable in the high skill/high challenge environments that flow requires. Fostering comfort with these complex environments from an early age – for example by carrying responsibility within their family or tackling challenging tasks in school – might help someone develop the flow personality (Nakamura and Csikszentmihalyi, 2009).
However, it's often difficult to tell what constitutes a flow experience and what doesn't. Flow should be viewed as a subjective state that varies in intensity: it's not an all-or-nothing phenomenon but rather a continuum, ranging from the mundane to the profound (Csikszentmihalyi, 1975/2000).
On one end of the range are microflow activities: small everyday behaviours – like tapping our foot or doodling on paper – that we exhibit on a daily basis, sometimes even without thinking. Graef (cited in Emerson, 1998) views them as facilitators of larger, dedicated activities. They provide structure to our lives and small goals to keep us busy – basically, they help make reality manageable.
Individuals who have been deprived of flow or microflow activities experience negative effects
The other end of the range is made up of deep flow / macroflow activities: intense, uninterrupted focus sessions – afterwards we're often able to recount ourselves being "in the zone". In these moments we truly lose our track of time and sense of self.
While flow experiences are positive, they exist on the verge of negative emotions. In order to experience flow, your skills need to be not just met, but stretched.
In their report of challenge in artistic flow experiences, Banfield (2018) details the struggle experienced by artists in producing an art piece. The struggle is the result of the uncertainty of a blank canvas and the possibility for failure, as they can't simply undo their brushstrokes — the artists' perceived skills don't match the challenge.
As the artists start working, proximal goals arise. The artists feel as if they're having a conversation with their work. Banfield calls this "a nonverbal negotation between equipment and materials".
It's actually this struggle and the negotiation that, in part, give the work its value and the artist its validation. In a way, moments of friction help highlight the moments of flow.
Flow in creativity is not merely the enjoyment of finding solutions, it is also the challenge of finding problems, and then finding solutions.
Another example is the user journey of the Dutch Railways (Van der Made and Van Hagen, 2017). There, boarding and exiting the train are considered an unchangeable annoyance – but it makes the train journey itself feel all the more peaceful and relaxing.
In order to visualise the skills/challenge balance, Csikszentmihalyi (1975/2000) developed the model of a "flow channel". The model was later adjusted to account for the progressive growth of skills during flow. When one operates in a "low skill – low challenge" context, the result will be apathy (Nakamura and Csikszentmihalyi, 2009).
The new model is centered around an individual's personal skill- and challenge levels, whatever those may be. Around it are eight sections denoting the various states, split into rings to denote the intensity of the experience (Csikszentmihalyi, 1997).
There are other models using only the four basic channels (flow-anxiety-relaxation-apathy), and yet other studies use a 16-channel model. This makes it difficult to compare different studies (Emerson, 1998).
As detailed before, flow is a highly subjective experience. This makes it somewhat difficult to measure objectively. Still, multiple measuring techniques have been developed.
The main problem with measuring flow is that any intervention will disrupt the flow experience (Emerson, 1998). As such, most measurements are done by having participants recall past experiences of flow. This is often done through retrospective interviews after an activity. Unfortunately, these technique might be subject to recall bias or other forms of response bias.
Other techniques aim to observe participants in the moment ("in situ"), as they are experiencing flow. One such method is through ethnographic observation. Another is through active participation in the same activity.
Practice-based sessions lasted approximately two hours and entailed both researcher and participant engaging in artistic practice while discussing that practice, enabling close observation of the phenomenon under investigation by participation in the lifeworld of the participant.
Another in situ method is the Experience Sampling Method (Csikszentmihalyi, 1975/2000), which uses electronic pagers to prompt participants - at random intervals in their day - to fill out a form describing their current activity and experience thereof.
While these techniques are closer to the actual experience, they are subject to observer bias and the Hawthorne effect.
Finally, there appear to be some developments in using physiological features such as electroencephalogram (EEG) activity (Katahira et al., 2018) or galvanic skin response (GSR) (Nacke and Lindley, 2008) to measure arousal and/or flow. Objective measure like this would be a great addition to the toolset, but the current ones are far from perfect. Physiological changes aren't the result of emotions, rather emotions are the result of the brain responding to physiological changes within a specific context, and these responses are not universal (Barrett, 2020). This makes it really difficult to assess flow from the outside.
The digital landscape has grown and matured in just a few short decades, and numerous best practices and design principles have been developed for the design of digital applications. Many of these principles affect the potential for flow state, and all blend into each other. To illustrate the state of digital tools, three trends will be discussed.
As mentioned, tools help you do something. There's little use in a tool that does a bad job helping you. This is why the value of a tool is often a result of the reliability or efficiency with which that tool functions.
This is similar for digital tools: they are focused on making the user feel productive by achieving their goals as efficiently as possible. Much of interface design has been concerned with presenting all the information a user could possibly need, along with all the actions they could possibly want to take.
But a mad dash towards the finish line isn't always the way to win the race.
Activities that look unproductive (or at best tangentially relevant to a job) may actually be crucial for long-term success.
Sometimes ideas take a while to develop, and innovation is hardly a straight line from cause to effect. The ability to experiment fosters the creation of proximal goals and flow (Banfield, 2018). Palmer (2020) addresses this problem in communication apps:
Our current software is too plain, based on a purely utilitarian exchange of information.
According to Palmer, software gives us cognitive presence but no human presence. It only allows us to express ourselves in data it can save, which inhibits our expressivity, emergent motivation, and thus flow.
Digital tools could use more play.
In game theory, a distinction is made between playing and gaming (Callois, cited in Deterding et al., 2011).
Whereas paidia (or “playing”) denotes a more free- form, expressive, improvisational, even “tumultuous” recombination of behaviors and meanings, ludus (or “gaming”) captures playing structured by rules and competitive strife toward goals.
Play is similar to flow experience: it is an autotelic (intrinsically rewarding) experience caused by emergent motivation through proximal goals (Prensky, 2001). A difference might be that play itself is considered relaxing, not flowy. Games are a formal set of rules and goals around a play experience, which add challenge (and the need for skills to match them) to the play experience. Games are designed to adapt these challenges to the player's skill level, either up-front with a difficulty selection menu, or on-the-fly depending on the player's current performance, resulting in flow.
In an attempt to capitalise on this, digital tools have started incorporating game elements such as achievements, levels, and leader boards. This practice is known as gamification (Deterding et al., 2011). Examples include Google Maps's "Local Guides" feature (Moore, 2019) and Todoist's "Karma" system (Todoist, n.d.).
But gamification isn't the magic path to flow that it might seem.
For one, gamification rewards the user for engagement. These rewards are a form of extrinsic motivation: the reason we engage with the app is because we want more rewards, not because we find using the app rewarding in itself. Not only is this form of motivation weaker than the aforementioned intrinsic motivation, it actually reduces intrinsic motivation (Lepper, Greene, and Nisbett, cited in Vohra, 2020).
Gamified tools focus almost exclusively on ludus – the rules and challenges of games, but do not implement paidia – the exploratory, free-form play experience (Alfrink; Deterding, cited in Deterding et al., 2011). The only place where gamification appears to work, is where the underlying system is already a game (Vohra, 2020). For other cases, such as digital tools, play should be a key consideration of the design process.
One way of doing this according to Vohra (2020) is by creating "software toys". An example they give is the search box in their email app, Superhuman. Its parsing is intelligent enough that people will start experimenting with it, "trying to break it". In doing so they will learn, through play, how to use the search box in new ways, and be more productive in the long run.
A good way of embedding play in a tool is through spatial interfaces. Humans experience the world spatially from birth, but most of our software doesn't use that built-up knowledge. Spatial interfaces use our understanding of 2D and 3D space to communicate how they function (Palmer, 2019), by simulating a world of "objects" (things to interact with) and "bodies" (representations of users) (Palmer, 2020).
A prime example of this is design tool Figma: it presents an infinite canvas that objects can be placed on. Objects can then be freely morphed and transformed with the mouse. Other people present in the document are represented through coloured cursors.
Spatial tools like this are more intuitive to use because of our natural understanding of space (Palmer, 2020). Because they allow for a larger amount of expressiveness, they are enablers of creativity and innovation.
In a more general sense, digital tools that allow for play cause their users to be intrinsically motivated as the interaction itself brings them joy and causes flow. It's important to remember that tools don't have to be devoid of play, and neither do they need to be games to be engaging.
A core principle of flow is knowing what to do next through proximal goals and emergent motivation. Modern software has increasingly tried to lower the bar of entry for users by predicting the likeliest next step through artificial intelligence (AI). AI can play multiple roles in digital tools (Van Bodegraven, 2019).
One is that of a curator. The most prevalent applications of AI are recommender systems (Ricci, Rokach and Shapira, 2011), particularly in social media (Twitter and Facebook's algorithmic feed (Cox, 2016)), e-commerce (Amazon's and Target's product recommendations based on buying behaviour), and entertainment (Spotify's Discover Weekly playlist (Popper, 2015) and Netflix's personalised recommendations (Chong, 2020; Netflix, 2016a; Netflix, 2016b)).
Digital tools incorporate recommender systems to surface contextual suggestions. Examples include Siri suggesting apps, Google Maps suggesting things to search for, and Microsoft Office's UI showing commands based on what you're currently doing with your document (Warren, 2020; Friedman, 2020).
These suggestions can improve flow as there's always a next action to take — but they have to be implemented carefully. For one, they need to be transparent in their functioning. Target's shopping recommendations were so accurate that it could predict a girl's pregnancy even before her father found out (Hill, 2012). When this creeped people out, rather than be more transparent about what its ads were based on, Target starting mixing up its ads to "look randomly selected".
In other cases, it might even be impossible for computers to judge which suggestion would be the best, particularly when it concerns subjective preferences.
I strongly believe that the definition of beauty is different for every person, everyone has a different perspective on that. There is no objective truth in creativity, and no objective measure of great photography.
Another potential issue is that of filter bubbles: when you only take actions based on algorithmic suggestions, and those actions feed back into the same algorithm, it'll be harder and harder for you to explore outside of your own bubble of knowledge (Pariser, 2011).
Lastly, recommender systems generally require a ton of information about the user they're optimising for. Capturing and saving this data is a potential privacy issue (Van Bodegraven, 2019).
Another role of AI could be that of an assistant. AI can perform tasks that humans are incapable of, or don't want to bother with. Examples of this include Photoshop's AI filters (Vincent, 2020) which modify a picture, or tools that outright create a picture for you (Adobe, 2017; NVIDIA, 2019). These AI features allow users to be more creative, even if they don't know which buttons to push. It lowers the challenge and skills required, and thus allows more users to enter flow.
A potential problem with entrusting AI to take actions (not just giving suggestions) is the lack of transparency and control since it's difficult to understand how the AI makes decisions (Van Bodegraven, 2019). And while AI itself isn't biased, the biases of its creators often slip into its functioning (Kantarci, 2020). In the long run, the more we delegate to AI, the further we remove ourselves from understanding how and why we do things (Van Bodegraven, 2019). We thus must be careful to design assistants that are powerful yet transparent and controllable.
Golden Krishna paints a picture of the state of interface design (2015). With the arrival of smartphones and marketing advertising millions of apps, everything could become an app. But the apps made were shallow, playing to easily satisfied but never-ending needs with endless feeds and micro-transactions.
The best minds of my generation are thinking about how to make people click ads. That sucks.
Moments of deep flow are free from distractions to the point where one loses sense of time and self. This can only occur during intense moments of focus within a loop of clear and unambiguous feedback.
The real problem with the interface is that it is an interface. Interfaces get in the way. I don’t want to focus my energies on an interface. I want to focus on the job ... I don’t want to think of myself as using a computer, I want to think of myself as doing my job.
The best interface, Krishna concludes (2015), is no interface at all.
The most profound technologies are those that disappear. They weave themselves into the fabric of everyday life until they are indistinguishable from it.
According to this "No UI" movement, software should work in the background as much as possible. Computers should work intelligently to adapt to individuals instead of prompting users for input, a concept Krishna (2015) calls "back pocket apps".
These experiences have the potential to be ubiquitous, continuous, and frictionless. All you have to do as a user is live your life, and the computer works around you. These are wonderful qualities for flow. But attaining that vision isn't easy: gauging human intention correctly is difficult, as exemplified by the errors voice assistants still seem to make. When these smart-but-minimalist UIs mess up, it's often frustrating that there isn't a manual override, or at least more granular controls.
As users, we crave convenience – especially when starting out with a tool we don't know. But "embracing typical process" is difficult to do when a user doesn't have a typical process that can be learned yet (a phenomenon known as the cold start problem (Bobadilla et al., 2012)). Eventually, we grow to need more control over our tools as we start to know what we're doing. The workflow we need might not be the one prescribed.
While more UI usually doesn't benefit immersion, the removal of UI isn't the end goal in itself either. Instead, the best interface is an interface you don't notice. But it doesn't have to be completely invisible to do that.
The ideal product design is the design you don't notice because you're so focused on the content that you don't notice there's something in between that's enabling that.
One that steers you in the right direction, challenges you in the right ways, all without you needing to ask anything of it. It needs to work around you in the same way "No UI" principles dictate, but doesn't have to remove itself outright. The best interface is never in your way, but still there if you need it.
Instead of asking people to modify their thoughts and actions around the arbitrary sandboxes of Apps, Mercury responds fluidly to the intentions of its user, alleviating the risk of interstitial friction that all multi-tool workflows carry.
Mercury's UI consists of modules: these are pieces of content that can be acted upon – emails, calendar items, locations. Instead of dealing with each module in the app they belong to, modules can be pieced together at will to create a flow — taken together, a flow is a way for you to accomplish a single task, like responding to an email and scheduling an appointment.
Flows can be grouped together (manually or automatically) into spaces: these are overarching intentions such as "review my inbox". Each flow in a space can be dealt with sequentially, and each flow contains only the needed modules to keep you focused on the task at hand.
Mercury is still very clearly a user interface (and a pretty one, at that). But instead of merely being a polished layer on top of a black box, Mercury is malleable by the user to fit around their workflow.
So far, this thesis has provided an overview of the current thinking in flow research and tool design. We've seen great examples of flow-inducing interfaces, as well as tools that make flow very difficult. Why do the good examples work so well? Have the bad examples deliberately blocked flow? What needs to be changed about them?
On the whole, it's unclear why some conventions work tremendously well in adding flow while others don't. The reason for this, I'd argue, is because there is no unified model for understanding how users are able to experience flow in an application.
While flow theory does not discriminate between flow in the physical- and digital world, most accounts and studies have been focused on physical tasks. Similarly, the theory on tools and our usage thereof has focused on the physical, down to the definition. The few studies that do study flow in digital tools have been focused on consumption and engagement, not creation and productivity.
On the other hand, the design practice has seemingly paid little attention to flow theory, or has at least had a difficult time implementing theory into practice. Market forces and artistic intents are often stronger than theoretical models. We've focused on what to create next, not on why we create something in the first place.
My job is to solve people’s problems, but as an industry we’ve gotten away from solving people’s problems. As an industry, we’ve gotten caught up in a globally evident technological impotence of me-too thinking that is taking us away from real innovation.
It's made our interfaces complicated, attention-grabbing, and sometimes even unethical. Meanwhile, the core way of interacting with digital tools has remained virtually unchanged from a decade ago — buttons, sliders, and a desktop-metaphor.
We shape our tools and thereafter our tools shape us.
Our interfaces are directly influenced by our design process. While we might talk a lot about flow theory, we don't actually know how to design for flow because there is no clear design process. The under-appreciation of a user's experience of flow leads to an underemphasis of causing flow in our digital tools.
Instead of relying on buzzwords and features, there is a need for a more concrete mental model for flow in digital tools.