TÍTULO X NORMAS COMPLEMENTARIAS DISPOSICIÓN COMPLEMENTARIA
SEGURIDAD INDUSTRIAL Y SALUD OCUPACIONAL
As illustrated in the key concepts in Chapter 2, the definitions around research have always been fuzzy. These definitions were built primarily around definitions that embed assumptions about the linear model of innovation, and have challenges when taken out of traditional laboratory contexts or applied to contemporary information-intensive inquiry. Some research inquiry carried out on fundamental and basic research about the social, behavioral, health, and public health aspects of individuals and society are obvious—others challenge traditional notions of what constitutes a research activity. This section explores the definitions offered by interviewees on what constituted research activities within their practices and firms, and the ways in which these offered definitions align and contrast to traditional definitions. This section also considers important nuances to research and information-intensive innovation within the private firm. These nuances are partially weighed by grappling with the surface studies of mere correlation compared to deeper inquiry that aims to understand the
fundamentals and “underlying why” in contemporary practice. These challenges and explorations on the boundaries of research illustrate the pressures on
traditional linear-derived definitions that constrain activities to particular contexts or assume limitations in motivation depending on the researcher.
There is rarely consensus in public statements or documentation or among practitioners on exactly which practices constitute R&D, or how their
organizations define and structure these activities. Interviewees were asked about their experiences and mental models about this boundary of research and
practice, including the distinction between UX (user experience) research and research that contributes to general knowledge.361 Many interviewees understood
that many of the practices within their firms or daily jobs involved gray areas: making inferences about users, writing algorithms based on user-generated data, designing new interfaces or hardware with user feedback, or conducting A/B style testing in real time. For the private sector, the distinction between product
development and research—or more generally R&D—remains blurry. As this
361 “Generalizable Knowledge” is a term frequently used in policy to determine what counts as
research. For example, if your findings are only interesting and applicable to your team or supervisor, then it would not be considered generalizable knowledge. However, if your research, say in UX, uncovers something about human biorhythms, the findings may be important for experts (or research) in that field, not to mention the general population. This blurry line is further discussed in this section.
section will put into context, this distinction has always been vague and
contested, yet understanding the contested points and boundary space can help frame what is included (or not) from additional discussions in this dissertation and beyond.
Contemporary Boundaries of Research
The boundaries of research, particularly as it is practiced within the tech industry, are placing new pressures on existing definitions of what counts as research. The introduction of user experience (UX) researchers and data
scientists present new questions into professional categories and classifications of industry inquiries and knowledge acquisition. Additionally, increasing numbers of PhDs are being hired by industry for their research skills and subject matter expertise, but for research ambiguous positions.
The Basis illustrative case (Chapter 5) introduced a discussion on the
differences between what the interviewees described as “business analytics” in contrast to the descriptions of the role of the research team within a sensing startup. Other interviewees frequently categorized academic style research, design (including UX, as well as some parts of product design), and business analytics articulated as three different areas of “research”—each with their own varying levels of rigor, relevance, and generalizability. Continuing to build off of the traditional definitions of R&D above, often practitioners interviewed used place (e.g., context of a lab or formal R&D team), the motivation of inquiry (e.g., to develop generalizable knowledge), and metrics used to evaluate success (e.g., peer reviewed publications) as elements of a definition. Often in practice,
however, these constructs were messy and operated in boundary challenging ways.
Interviewees often used methodology, scientific rigor, and the qualifications of those on the team as qualities and elements to reflect on differences in research outputs within the firm.
“I don’t think there’s anything flagrantly wrong with the other groups and their research, but we have a very academic approach from how we do things and think it through. [We do this by] making sure the study design is as objective as
possible, [considering] where the participants are at, and what they are representative of and so on. It’s just not the same kind of rigor because other groups tend to be scrappier on ‘let’s get data and just move with it,’ and at the end of the day they are just a product group….”
In grappling specifically with UX research, there were mixed reflections. For instance, some interviewees stated that it mattered what the UX team found, and how actionable the findings were outside of the team. This implies that it is
possible for a UX researcher, who does not traditionally produce generalizable knowledge, to begin depending on the subject matter at hand—meaning UX research cannot be wholly cast as non-research but rather operates at the
boundaries. One of the missions of UX teams is to find where users get stuck in using the platform or device, and improving the usability and experience of the user. If that friction was caused by a deeper underlying behavior or need within the app, there was sometimes crossover in teams to handoff developing and understanding these capabilities. One engineer from Basis reflected on this handoff:
“Studies about user experience. That’s, I think, probably…probably more on the business intelligence side. They would see where things are going well, where things are going not so well. Then, if that information was actionable, by the algorithms team, then we would start researching ideas on how to improve it.”
This perspective introduces UX as an important gateway to other research (either by the same individuals or a different team) and almost like an
exploratory or feasibility study to prepare for more formal inquiry. Many other interviewees talked about the motivation for conducting the user experience. If the study was narrowly focused and very applied to fixing one particular
functional problem, practitioners classified UX as less of a traditional research activity. This directly relates to the traditional definitions, where the formal research activities depends on the goal and generalizability of the findings.
Interviewee A: “It feels like UX research tends to be a little bit more short lived and singularly focused, whereas research with a big R is a bit more longitudinal. It’s sort of, yeah, I would say it's longer lived and it's more rigorous in way, and reproducible.”
Interviewee B: “UX research is more connected to, there's a business opportunity and we're building a thing. And that can start at any point, right? From really early research about the [customers], but it's product oriented, or output
oriented, whatever that output is, and big R research is being much, much more exploratory and the output is the discovery itself.
Other interviewees added nuance, that exploratory UX work could lead to more fundamental investigations, that potentially opened up larger questions.
“I would tend to lose sight of this because my training is in people, and sociology and such, but like, there is also the test of like, can we build this thing at all? Can the technology be made to do the thing we're trying to do? Which is maybe a third kind of, half uppercase R. Whatever you want to call that.”
firm led to some broader underlying questions about human behavior and technological limits, which they found to fall between the two classifications of research and non-research.
Many practitioners interviewed also thought research, in its most pure sense, was not for any applied knowledge—thus framing only within the definition of traditional “basic research” classifications, even in technology fields that were from the onset of contemporary research heavily focused on applied uses. This framing focuses only on the goal of the activity.
“[Formal research is] without like, a goal of how you would actually apply that knowledge, it's more …it's more for the discovery and the knowledge.”
Similarly, another interviewee underscored the importance academic style outputs—thus further reinforcing the definition linkage to the goal. One
interviewee who had worked within UX at a formal lab commented: “It's for the knowledge and the output. I mean, the measurement of how successful people were in that role was how many papers you're publishing. So, it really was like you're a professor essentially.”
Location and designation within the company mattered to respondents, thus relying on institutional designations and professional contexts to define the underlying activity. For instance, Interviewee A referenced above made an immediate disclaimer after their statement about UX research being more short lived and singularly focused: “Unless of course like you're a UX researcher in a research lab, like at Microsoft or Intel or something.” This respondent overrode their previous caveat by placing the institutional context as the most important designation in determining what constituted a “formal” research activity.
This perspective was echoed by other interviewees. In these cases, the location (or institutional context) under these formalized institutions trumped, for many practitioners, any other discussion of research motivation, applicability of
findings, methodological rigor, etc. If the UX activity was done at these formalized research places, then it must be research.
Definitions reflected by the practitioners included the ways in which data collected about users were used, not just the type of data. For instance, one engineer provided the following classification in reflecting how his own team sectioned data for algorithms and research and how it was used by others like those on the UX team within the startup.
“Well, I guess there’s maybe three buckets. There’s the sensors, which we’re passively monitoring. Then, the data that we use—the derivative data, like were you sleeping and what’s your heart rate. That’s the sensor data. The user inputs might be self-login. Some apps, you can add activities. Then, there’s, I’d say, user or almost usage data, which is where are people spending time in the app, what buttons are they pressing, where are they getting hung up. That latter one is definitely, for me, [related to] product testing, because that’s a lot of how the product designers really critique their designs. They wanna see where people are
getting hung up, what’s confusing, where people are spending time in the app. Then, that drives product iteration.“
Not every practitioner I spoke with reflected the canonical arguments that “basic” research was a “pure” line of inquiry. One UX researcher, in thinking aloud about the differences and boundaries between UX and traditional research pushed back on the idea that industry UX was always biased because it was oriented toward a product or articulated goal. This researcher pointed out in her reflection that both industry UX and university research—despite the
protestations of academics—are both money (i.e., grant or profit) oriented, and that both are aimed at finding something that worked.
“That also like reminds me things that kind of happen in the sciences a lot where you are looking for a gap because it presents a grant opportunity, and so
everything is still, in its own way, motivated by money. So even for like the UX research, there's a clear business opportunity and it's got to be stated up front like, 'Okay, we want people to shop more so we're going to do this research.’ Whereas even the academic research guy has a cloud that may not be right on top of you, but it's in the near horizon… Science is supposed to be tangible and it's supposed to produce direct output , then there's kind of this, this need to [make it] work because this will get us funding, or this will get us an invention or something. So that's kind of like, that's always been like a weird issue to me, research.”
Many industry UX interviewees, or interviewees who worked with UX teams, reflected that the culture of the firm directly influenced the style of approach, and thus the resemblance of UX to more traditional research. This often had to do with the way activities were labeled, or expressed goals of the activities either in the formation of questions or the framing of outputs.
“It's not necessarily a number of studies, even though like, different leadership has different interest and investment in promoting a culture of testing, and
learning, and iterating.”
Many interviewees also felt frustration that since they worked between the boundaries of pure research and simple product testing, that they were frustrated they did not have more clear ways to evaluate success on their jobs clearly for the management.
There may not be a clear line between UX and formalized research, but
characteristics here including professional organizational context, goal of inquiry, flexibility to iterate and use rigorous methods all define the boundary space
between these two concepts. A formal line is not necessary to begin to consider where UX activities may bleed into research, and should be considered part of knowledge production within firms.
Correlations and Asking “Why?”
Several interviewees reflected more generally on a key difference between academic research and similar activities within industry. One of the key
components reflected was the drive industry practitioners had to achieve a result (e.g., click the button), identify a pattern (e.g., blue pill achieves more purchases than red pill), or optimize prediction without understanding the underlying and deeper questions of “why.” The “why” sometimes corresponds to deeper human behaviors or underlying causes in why reaction was occurring. In industry,
particularly in fields and jobs generally related to data science, the emphasis was on prediction and performance, not deeper understanding. The questions these data scientists were trying to answer stopped at achieving a practical result.
“I feel like I'm really old school, like I'm just like the old school academic there. But I do know that, like I was saying, at Intel Labs there are like, and probably at IBM too, there's like groups of people who they bring them in who are really trained scientists who are adding a lot of rigor to how they are thinking about the future of technology and the relevance of where it fits. The folks that I worked with at Intel would be great to talk to about that because they're doing really interesting research that's applicable about kind of probing into the whys to understand the cultural context for technology because if they're gonna make a new thing they need to know really what people want this and where it's going to fit, and they're doing real research. But what I've seen in kind of this more smaller scale setting is very parochial kind of just focused on optimization or something. Like, making a repetitive task faster or automated or improving the performance of an ad campaign and all that kind of stuff. But not really invested in knowledge advancement or theory testing or solving long standing
conundrums of the human condition or something like that.”
This interviewee also commented: “There's no thinking, there's no
understanding and just getting to your point about it. The research that goes into machine learning is generally focused on prediction and performance not an understanding. So, there isn't that kind of that mindset of inquiry about, like, why is this model doing that? Why is it predicting that? And is it something we want to be doing?”
Understanding the deeper and sometimes more philosophical “whys” is not without some drawbacks, as the interviewee noted.
“Yeah, and it's funny now because I've been [working in industry] and now, when I come here and go to a seminar, I'm just like, ‘Oh my God. You guys are so in the weeds.’ It just feels so esoteric and so like, ‘What are you doing?’ In some of the ecology seminars I'm like, "Why are you doing that?" You know, like, where is the ... And it's kind of cool but because it seems like there's a lot where you can
tell that it's science for a career so I just need to find this little niche and publish the least, you know. Like, just get as many papers as I can out and I don't care how kind of weirdly just sort of narrow it is.”362
An engineer, when talking about how they use data on their own research team within a startup, expressed the same difference between straight forward analysis and deeper inferences.
“Inferring user behavior habits, that one’s like—for me, it’s both, very much so, in that, if we know—for example, you’re like I wanna go out and detect slang, or walking, that’s something that we know is doable. You just go out, get the data, and develop it. If it’s more of this higher-level—or, I guess, more abstract, like how are people spending their times, we don’t know necessarily what they’re doing, but let’s do something on—let’s do something with activity that no one’s really done before, but maybe it doesn’t work out. That would be very much on the research side, if that makes sense.”
Interestingly, James Wang of Lioness had previous experience working at a hedge fund, where the whys underlying market correlations mattered a lot, and was emphasized by the firm. He reflected during his interview on the need to understand the deeper root causes of identified correlations, and the ways in which Lioness requires more correlation than a marketing company, but still less demand on resolution on the underlying causes.
“Which is interesting, and this is now going a little farther afield. I think it's an interesting comparison. For example, at Bridgewater, the hedge fund that I worked at before, we were very concerned about root causes. The reason is because if you didn't actually find the correct root cause, guess what? You might actually have this signal. You might have this thing that you expected would work suddenly snap back in your face and oh it didn't work at all because something, the correlation we thought was true will become false in a different market environment or something.
This is far closer to what we do, just because of the nature of the consumers that we're with, or looks far closer to what a lot of marketing companies have in terms of their attitude towards their marketing research. They're still pretty rigorous in terms of it. They still try to do as much as they can to try to figure out why's, but they're fine with correlational views. For the most part, in terms of consumer environments, it works. The problem with trying to do that in markets is they
362 This interviewee quickly went on to assure me this particular study was not one of those useless