I close by taking a step back, shifting my lens away from discussing data engineers searching at work, and consider what lessons and provocations my findings offer for search more broadly.

Web search engines have beengranted too much power ¹¹⁵ and searchers both claim and are assigned too much responsibility. The search box, SERP, and the results navigated are a small slice of web searching. Web search is made to work, where it is, not only by the technical mechanisms (the indexes, algorithms, and designs) and social construction (articulations or imaginaries) of the search engine—and not just by the searcher themself or the websites seeking to be found—but by interactions among the various components in the larger systems and the contexts in which search is situated.

Success in using search matters greatly, both for data engineers and for society broadly. The relatively tacit modes of sharing knowledge about productive use of search in data engineering work have disparate implications for historically marginalized people. The do-it-yourself image portrayed by data engineers hides a deep web of taken-for-granted dependence on the knowledge of others.

Below I will present two further arguments developed across the chapters, then briefly review my core findings and arguments with attention to applications beyond data engineering, highlight the importance of my findings to our understanding of search and their broader implications, and end with an appeal for imagining search differently.

Two further arguments: search envelopes and gatekeeping in data engineering

While web searching for data engineering is generally put to effective use, I also make two arguments about limitations. Here I will organize concerns introduced throughout the chapters. First, the effective use of web search by data engineers cannot be expected to be as stable for searches outside the structuring provided by the data engineering work practices. Second, the maintenance of search as solitary and backstage reproduces uneven access to community norms and knowledge, limiting people’s ability to make effective use of search (and gatekeep in a way that is at odds with efforts at changing the demographics of technical professions).

First, the limits to someone’s or some community’s success in searching depends on how resources around them are aligned or marshaled towards particular sorts of learning and doing. The effective search envelope is dependent on the knowledge of other people and artifacts. “Search envelope” refers to the operating envelope or performance envelope of the data engineer’s model of search and the surrounding system that supports the searching.¹¹⁶ Without the right search inputs, relevant evaluation, and environment for acting on search, searchers will struggle (or leave search sessions misinformed). While generally successful for their purposes, the effectiveness of data engineering search practices is limited by the ambiguity of the search confessions, the taken-for-grantedness of web search and the occupational, professional, and technical components supporting it, and firms’ hands off approach to both search repair and responsibility for searching.

Efforts to improve search exclusively or primarily through improving search literacy of individuals, by building greater technical expertise in the mechanisms of search, are missing important ways in which search is made to work in contexts. My findings reveal that web search in data engineering is constrained by the imprecision of the confessions, situated searching supported and scaffolded by the extensions of search (in the query generation, space for evaluation, and decoupling), the repair work, and a possessive and shame-inflected approach to knowledge and ignorance. The limitation on successful searching is akin to an individual attempting to search or learn outside their “zone of proximal development” and without scaffolding from experts and the environment. This is related to notions of a firms’ “absorptive capacity” (Cohen & Levinthal, 1990, Roberts et al., 2012) , their “sensing routines” (Carlo et al., 2012, p. 870) , or “range of adaptive behavior” (Woods, 2018, p. 435).

Data engineers and their organizations can effectively rely on searching because it is supported by occupational, professional, and technical components of their work practices. Their success is not the result of individuals’ sophisticated technical knowledge of the mechanisms of search. Data engineers apply their domain expertise and draw from code, exception messages, and interactions with colleagues in developing search queries. A single data engineer learning an error-prone code pattern from a web search is likely to learn of the problem in their attempts to run code, in code review, or in testing prototypes.

Second, not all of the data engineers are fully brought into success of the searching practices. As researchers look at factors that push people out of technology work, this analysis of the situated searching experiences of data engineers presents an interesting case study of interactions concerning people’s status and inclusion within a workplace. I discuss how people describe the effects of a fear of being mistreated and misjudged because of how their peripheral position within the system shapes their performance and perceptions of their knowledge and ability, their skill and thus their responsibility. The general hands-off approach of management towards searching, informality around talk about web search, and lack of technocratization of search (intentional application of technique to influence search activity) produces “a way of masking power” (Freeman, 2013, p. 232) and maintains the “fiction of technological meritocracy” (Hicks, 2017, p. 16) , providing illegitimate cover for reproductions of hierarchy.

A data engineer’s expertise is a shared accomplishment. Data engineers’ ability to function as knowing experts—to know and to be seen as knowing is a product of their situated searching with the support of occupational, professional, and technical components of their work practices. Power relations are maintained, in part, by the invisibility of the searching practices—the engagements of the system taken-for-granted—and the maintenance of expertise as an individual possession (rather than composed in networks (Cambrosio et al., 2013, Eyal, 2019) ). These web search practices can be a barrier to new, and particularly already marginalized, data engineers’ full participation in their workplace. These findings provide a partial “account[ing] for the normalization and production of systematic advantage” (Hoffmann, 2019, p. 910) and disrupt notions of “exclusive forms of technical expertise” (Hoffmann, 2021, p. 11).

Lessons for searching and further research

This research shows how data engineers have made web search work for them, with lessons for research and the teaching and design of search more broadly, and revealing in the process limitations to their organizational learning and inclusion. My study of a discrete group of people, data engineers working in companies across the United States, provides lessons for our understanding of search that are applicable to searching in other domains and communities. I conducted interviews and performed document analysis with the two analytical lenses of Handoff and LPP to see how search is learned and used and how thathow impinges on core societal values—namely responsibility, privacy, and fairness. Iterative comparison to literature and examples of searching in other domains and cases shaped my attention and analysis. This site—with data engineers technically sophisticated, well-resourced, and reliant on search—was selected in order to develop broadly applicable lessons. While it could be framed as a special case of broader interactional and organizational processes, looking back to work from Goffman (1956) and Thompson (1967) , I focus here on the lessons and inspirations most closely related to the design and use of web search.

I will quickly review the core ideas, the findings and arguments, and then reflect on key takeaways.

This research tells a straightforward story of data engineers’ success in making use of web search for work (though with a couple clear caveats and concerns raised along the way). It starts with the observation that data engineers are reliant on web search for their work, seemingly successfully, and present a potential ‘best case’ study for exploring the role of technical knowledge of web search mechanisms. Interviews reveal there is limited explicit instruction, discussion, demonstration, or collaboration in the moments of web search in data engineering. But search confessions legitimate their searching, shape norms of use, and direct others to also rely on web search. Data engineers’ success in search is not because they know more about the technical mechanisms of search, but because their work tools and practices (and domain expertise) make search work for their work purposes (supporting query formulation, evaluation of results, and decoupling performance from search automation bias). The occupational, professional, and technical components, as extensions of web search, provide sites and activities for new data engineers to gradually increase participation in the search work of data engineers. Search repair practices provide data engineers both additionaltalk about search, further legitimating it within their work, and opportunity for the learning data engineers to participate in extensions of web searching. The search repair practices constitute articulation work necessary to support such heavy reliance on web search. With firms delegating responsibility for searching down to individuals as a strategic approach to uncertainty, individual data engineers identify themselves as responsible for their web searching. The firm (pursuing only the delegation and no further ownership of search) nor the data engineers have intentionally applied technique to influence the search practices themselves. Successful data engineering use of web search is constrained to the imprecision of the confessions, situated searching supported and scaffolded by the extensions of search, the repair work, and a possessive and shame-inflected approach to knowledge and ignorance. Not all of the data engineers are fully brought in to participate in the success of the searching practices.

Below I identify five takeaways, and for each a key observation and provocations for future research.

1. Web search in data engineering is continually re-legitimated, in this case, through talk about search—search confessions and search repair work.

Identifying similar legitimation work, or resistance and denial, to web search in different situations will provide entry points for understanding relations of power and constructions of learning, knowledge, and expertise within a situated practice—beyond the questions of web search practices themselves.

Such examinations may include looking at participant interventions to shift the legitimation as well as the modes of engagement (targeting, for instance, perceptions of constraint, affordance, shame, or celebration). How are search confessions and search memes (“just google it”, “google is your friend”, “LMGTFY”, “google knows everything until you have an assignment”) enrolled in other settings? Search directives? Are there settings where web search has reached something closer to closure (Bijker et al., 1993) ? Such legitimation work is more formal in some settings (such as in proscriptions of searching the web in classrooms (Haider, 2017) ), but may appear in jokes, memes, or something similar to confessions elsewhere. In healthcare related contexts there is a meme that patients find printed on coffee mugs and posters: “Please Don’t Confuse Your Google Search With My Medical Degree”. The legitimation (or not) of web searching may not be discursive, but dictated in relations of access (Burrell, 2018, Haider, 2017, Robinson, 2009, Sundin et al., 2017). What are the calls for memorization rather than reliance on web search? Who is using their own search engine? Legitimation of googling in school work is heterogeneous (interviews expressed differing experiences). Does legitimation of web search reliance in schooling in some fields map onto the use of web search in the workplace? How is the legitimation in these various settings taken up by or against different people within the setting?

2. Web search is extended beyond the search box and the SERP.

This is not a new sort of claim, but following it provided visibility into the role of knowledge and space for participatory learning in this case. Identifying the extensions of search in different search situations may point to potential reconfigurations to reduce the dependence on individual search performance.

Are there search practices that distribute more aspects of searching (than query formulation and results evaluation) to other systems and people? How are seeds disseminated and found in other settings? Are there other search settings where evaluation of results can be done so decoupled from the search and the performance effects? What types of searches, in what settings provide the least space for evaluation and decoupling? How are searches for the various “Your Money Your Life” topics—topics that have “a high risk of harm because content about these topics could significantly impact the health, financial stability, or safety of people, or the welfare or well-being of society” (Google, 2022) —extended in different settings? How do different settings provide impetuses for search? Can seeds be disseminated as an intervention?

3. Successful use of web search did not hinge on personal knowledge of the technical mechanisms of web search, in this case.

Identifying the extent to which the successful use of web search in different situations and search purposes depends on varying degrees of domain knowledge or personal knowledge of the mechanisms of search will provide entry points for understanding the embeddedness of expertise and decoupling.

How does this finding change if the non-work related searches of this group of people are also scrutinized? Under what conditions does personal knowledge of the technical mechanisms matter significantly in other cases of web search, search, or the use of other information technologies? How do we identify the boundary conditions? What specific advantages might knowledge of the technical mechanisms of web search provide in different situations? Do very high levels of domain or search mechanism expertise counteract search automation bias? How does frequency, variety, or urgency of searching interact with these questions? Social media is sometimes used as a substitute or complement for web search (Oeldorf-Hirsch et al., 2014, Shah, 2017) and the use of question-and-answer sites like Stack Overflow, Quora, and some subreddits (Gilbert, 2018) have been characterized as social search or “asymmetric collaborative information seeking” (Morris, 2013, Morris & Teevan, 2009). How is this used, not as a substitute or comparison, but to formulate or evaluate web searches?

4. Web search is entangled with notions of responsibility, both credit and blame, for (and possession of) knowledge, in this case.

The hiddenness of searching to protect people’s status as knowing individuals, concealing their ignorance or the resources they employ to learn is found in many domains of searching. How could we encourage more sharing of searches while retaining the value of intimacy for learning, retaining the safety to search within cultures that still penalize the learner?

Might we draw on contextual integrity (Nissenbaum, 2011a) ? Can we find examples where such sharing is practiced and effective? Is web search treated differently by practitioners who adopt lessons from human factors and safety science (blameless postmortems, no single cause of failure)? How does the TIL (Today I Learned) movement treat responsibility in search? How does the use of school or workplace “no stupid questions” channels treat responsibility in search? Are there elements of open source coding movements, with the tension between generalized reciprocity and individualized responsibility, like those studied by Weber (2004) , Coleman (2012) , Dunbar-Hester (2020) that have varying treatments of responsibility for knowledge? Koonin (2019) ’s “Everything I googled in a week as a professional software engineer” is widely shared, has her professional experience, and that of those who also publicly shared (or share) their searching, shifted? How do teachers or livestreamers who code and search live discuss the “obligation to know” (Reagle, 2016) and the assignment of credit and blame for knowledge? Tripodi (2022b) describes a “do-it-yourself” approach to search where propagandists appear to successfully convince people they need to think for themselves and that web searching fact-checking is the legitimate approach to that, all while feeding them search queries that constrain what they might find.¹¹⁷ Does the mantle for searching that these people have taken on still maintain a backstage for the actual searching activity, or is it proudly shared?

5. Technocratization of web search, the intentional application of technique to influence search activity, did not make an appearance, in this case, but it could be pursued in multiple forms with varying effects.

Web search conceptions, activity, and the tools themselves are plastic and different occupational or organizational factors may produce very different findings and futures.

While the search activity itself of the data engineers I interviewed were neither surveilled nor intentionally scaffolded, there are tools for this, as well as search engines and tools that may be substitutes for general-purpose search engines. This includes tools to facilitate memorization rather than searching, to substitute for a subset of searches. While it may be easy to identify examples of workers who do have their searching activity formally surveilled, restricted, or nudged by management, how do they resist or adopt these constraints/affordances? How are people using large language models to influence different aspects of search activity?¹¹⁸ Search engine optimization experts use a variety of tools to better understand searching behavior of potential customers, do they adapt those tools to their own everyday or YMYL searching activity? How can searchers, and designers supporting them, make or modify tools to better see, distribute, and accomplish the always unfinished and context-dependent job of the search user interface?¹¹⁹

to aid users in the expression of their information needs, in the formulation of their queries, in the understanding of their search results, and in keeping track of the progress of their information seeking efforts.

Why search matters

How we search and how we think about it shapes not only what we learn but how we learn with and from others. Questions about relying on web search in the workplace are not about deskilling or even reskilling so much as about what skills are, who is allowed or seen to have skills, and “the political disenfranchisement and dehumanization of those people who are categorized as unskilled” (Iskander, 2021, p. 256).

People will attempt to turn, in part, to web search—or some transformed variant—to understand their changing world, including climate emergencies and resettlement, wars and their rumors, future pandemics and new vaccines, changes in schooling and healthcare, increasing inequality, and myriad new technologies from robotics and gene-editing to artificial intelligence and green technologies. Web search will continue to be a battleground where democracy is defined and practiced. These and other “looming disequilibria” (Weber, 2019, pp. 15–17)¹²⁰ prompt potentially life-altering queries that people will navigate within their systems of searching, for better or worse. Who will make successful use of search? How?

Imagining searching

In this research I have pursued “provocative generalizability”, I “attempt[] to move [my]] findings toward that which is not yet imagined, not yet in practice, not yet in sight. [ . . .] rather than only understanding (or naturalizing) what is” (Fine, 2006, p. 100). This is “a normative orientation” (Liboiron, 2021, p. 154). The lessons from this dissertation can be put towards defamiliarizing (Bell et al., 2005) web search and joining others as they “make, unmake and remake the search engine” (Sundin et al., 2017) , “imagin[ing] search with a variety of other possibilities” (Noble, 2018, p. 180).

Vaidhyanathan (2011) concludes his exploration ofGooglization with a section titled “Imagining a Better Way”, writing that “[t]he question is not whether Google treats us well but whether this is best we can do.” We can make new search engines and more people can make better use of search through their practices and reconfigured contexts.

We can find ways to clearly legitimate effective web search practices, celebrating searchers rather than stigmatizing them. We can learn to distinguish effective search practices from those that are manipulated or poorly modeled and likely to misinform or fail to inform. While we must seek knowledge of the mechanisms of web search engines in order to reshape or replace them, we can find places to search around the opacity. We could share habits and practices that are not constrained by lack of transparency on the part of the decisions of commercial web search engine companies or inherent in the systems they build. We can focus on building the knowledge for effective searching into our practices, tools, and environments. We can work on mobilizing and recognizing effective search seeds in different domains. We could focus on developing and calibrating our individual and collective ability to evaluate search results and our results-of-search. We can look for configurations of components that let us decouple from search automation bias. We might see more of our interactions as spaces where we participate in formulating and evaluating searches with others. We can spread the practices for search repair that connect and encourage people rather than cut them apart and tear them down. We can see the extensions of search and find or fashion our own techniques to scaffold our searching and refashion our search practices. We can address search gaps in ways apart from turning to automation. We can determine how to share our search activities in ways that are appropriately sensitive to the relations between people and their goals in different contexts. We can make the extensions and effects of search more visible. We can make talking about search less shameful. We can find more ways to search together. We can recognize how search is a shared performance and can be a shared responsibility.

Bibliography