Workgraph: personal focus vs. interruption for engineers at Meta

Using Logs Data to Identify When Software Engineers Experience Flow or Focused Work

“…Additionally, while other behavioral methods have been able to begin to isolate states of flow [17,33,34], it is important to note that the current approach may represent the first approach to doing so unobtrusively and without requiring further self-report data. Similarly, prior work that has described patterns of fragmentation and focused behaviors have not included any type of quantitative measure for this phenomenon and relied solely on intuitions around what behaviors should and should not count as focus [2,24,35]. The current research introduces a behavioral metric that quantifies focus without requiring these assumptions and instead uses a data-driven approach to identify behaviors that are related, which, as we show through the validation analyses, reveals how focused these behaviors are.…”

“…While leveraging fine-grained IDE data is useful to describe patterns of behavior that have been interpreted as "focus, " at no point were the research participants asked to characterize their actual experiences (i.e., we do not know if the patterns described in the IDE data actually felt like focus or flow). More recently, Chen et al [2] developed a measure of focus that leverages logs from a number of work tools (e.g., IDE, chat, internal wiki) by grouping together interactions that do not have interruptions (e.g., a chat message) and present descriptive statistics about how much time engineers spend in focus, as well as what factors impact this time. Similar to the work above, this approach describes focus based on assumptions about which behaviors characterize focus (e.g., using an IDE) and which behaviors do not (e.g., chatting with a colleague).…”

“…While the current work was able to leverage fine-grained behavioral data from a number of tools that developers use to complete their work, the described approach could be applied to other sources of behavioral logs data that capture human interaction with a technology. One of the potential gaps in previous work that has assessed how focused engineers are based on IDE and tool data is that descriptions of focus are based entirely on mental models and hypotheses rather than quantification of the data [2,24,35]. This method could be applied to these data sources in order to better characterize which behaviors actually reflect focused work and, in the right scenarios, flow.…”