Trapped in the 20th Century?
Why Models of Organizational Learning, Knowledge, and Capabilities Do Not Fit Bio-
Pharmaceuticals, and What to Do About That
Deborah Dougherty, Professor
(essay forthcoming in Journal of Management Learning, based on Keynote Talk given at
Organization Learning and Knowledge Conference, March 2006)
This essay highlights three traps that keep academic models for knowledge and
innovation management trapped in the 20th century, and suggests ways to rethink these models to
address 21st century innovation challenges. The traps: privileging technology over other
knowledge systems that are involved in R&D and in innovation, presuming that innovation work
is decomposable, and presuming that products are scalable. I stumbled across these traps in a
study of product innovation in bio-pharmaceuticals (focusing on the first six of a twelve –
sixteen year process that deals with discovering new drug candidates). After more than a year of
grounded theory building efforts, I see patterns lurking in the data but cannot grab them
conceptually. My own limits are no doubt a factor, but this confusion also reflects lacuna in our
field: existing models help only by what they do not cover. More than that, reaching into the
theoretical plane from the empirical plane of drug discovery reveals enormous voids, not just
gaps. Existing models, however implicitly, focus on the special case of 20th century technology
(e.g., autos, computers, silicon wafers, chemicals), and emphasize linear, decomposable,
scalable, and path dependent processes that obey the laws of physics. Many 21st century
innovation challenges involve non-linear, non-decomposable, non-scalable, and pathless
activities that obey the “laws” of life sciences and social sciences. Among 21st century
challenges are health care, various combinations of “bio” and “nano” with other knowledge
systems, ecologies and climate changes, and terrorist and intelligence systems.
My ideas are not new, since others raise similar problems and offer frameworks for
thinking them through (e.g., Calloun, Carlile, Dosi, Latour, Nelson, Nightingale, Pavitt to
mention a few). But the fact that almost all research presentations and publications are still
premised on these traps shows that the community at large does not appreciate these problems
with our models. I highlight connections among problems, ground the problems in the
particulars of drug discovery to bring them alive, and indicate some ways to rethink our work for
the 21st century.
Techno-hype? The idea that academic models overemphasize technology, even defining
R&D as a matter of technology, came to mind because the bio-pharmaceutical industry has done
so as well. Most big pharmaceutical firms have invested billions each into mega technologies
such as rational drug design, high through-put screening, combinatory chemistry, imaging
techniques, or genomics. Industry participants seem to have treated drug discovery as a
technology problem: bring in more machinery, devices, automation, assays, and other scale-ups
to do more things faster. However, pharmaceutical R&D productivity has declined despite this
massive infusion of technologies, funding, and patents (Economist 2005; Nightingale and Martin
2004). Most people interviewed are figuring out that they must also do the necessary science,
and that technology is not a substitute for science.
Their challenge, and ours, is how to go beyond the management of technology. While
our literature contains general discussions of differences between science and technology, most
empirical studies focus on technology, as if R&D and technology are synonymous. Society in
general and management academics in particular seem to also be caught up in techno-hype.
Popular magazines have pages and pages describing fabulous new techniques, devices,
approaches, testing regimes, or instruments that will revolutionize health care, especially drug
discovery. But there is no discussion of how all these fabulous gadgets will fit into the
underlying drug discovery process, or indeed what that process entails in the first place.
Academics study biotechs (or their patents) to explore agglomeration, spillovers, knowledge and
learning, and alliance network structures among other topics, but rarely address the fact that this
sector loses money every year, that the sector does not exist apart from “big pharma,” and that
many “biotechs” do nothing of the kind, being engaged instead in traditional small molecule
drug discovery or the creation of discovery technologies. I wonder about the validity of all the
research that is based on biotechnology and pharmaceuticals, since the research ignores the very
different innovation processes in this sector.
Techno-hype limits knowledge and innovation management by glossing over the
differences between technology versus science, blinding us to the fact of technology‟s blind
search, and blackboxing the nature of the knowledge involved. First, the work of science may be
very different from the work of technology, so managing science work like technology work
may in fact inhibit the science. However, despite the vast literature on the differences between
science and technology (see Allen, Bunge, Cardinal, Nightingale, Pavitt, Vincenti for
elaborations), prominent scholars tell me that there is no difference, and researchers who have
tried to distinguish the two tell me that reviewers have balked at such a notion – we may have a
strong if unstated “techno-hype” paradigm hanging over our heads!
To review the general differences, technologists seek to create useful products or
processes, often without explaining why those results are achieved. Technologists begin with
known outcomes or the desired functionality, and reach back into the complex space of the
phenomenon for unknown starting conditions that can achieve them. They work by weeding out
alternatives until functionalities are achieved, and by pulling ideas in to specific answers.
Scientists, however, proceed in an opposite way. Scientists seek to create understandings, and
reach forward from known starting conditions (theory) to search for unknown outcomes. They
work by weeding “in,” not “out,” and by opening up to general questions rather than pulling in to
particular answers. Precisely because of these qualitative differences, increased iterations
between science and technology can significantly benefit both. However, we cannot leverage the
complementarities if we continue to treat innovation as if it were technology alone. And we
cannot go beyond the generic differences until we delve into the different kinds of both
technology and science that may be involved in different industries. Finally, we cannot get very
far until we recognize that academic science may be different than industrial science.
The second critical limit of techno-hype arises from the first. Technology search is blind
since it seeks to achieve certain outcomes without explanation. But because we have such
powerful search engines (computers, assays, genomics) we generate vast amounts of information
that so far has little use. Nightingale (2004) suggests that there are more possible drug
compounds than there are particles in the universe, and there are thousands of proteins in the
human body that interact with one another in unique ways. Thus, technology search alone
cannot work in bio-pharmaceuticals, and people working on drug discovery are now figuring out
limits of blind search the hard way – and without much help from us.
Third, drug discovery involves life sciences, where knowledge is largely descriptive and
applies in a limited way beyond a situation, and where explanations must integrate several levels
of reality. But the nature of “knowledge” is typically black-boxed via simplistic abstractions
such as tacit versus codified, complex versus simple, or even exploit versus explore (this industry
is exploiting a fundamentally exploratory innovation process, so this popular dichotomy seems
useless). There are no architectures, platforms or even learning curves, and limited knowledge
accumulation because of the situated complexities. The underlying processes of innovation
depend on the number and types of knowledge systems that are involved, the extent to which the
knowledge systems are mutually constituted, and the subject matter. We need to particularize
our models to reflect these profound differences in knowledge, and dig down to unique aspects
of different subject matters to look for underlying dynamics that might generalize.
Important work in our field therefore is to understand R&D as a system of diverse
knowledge systems, to rethink the basic innovation dynamics to include these very different
processes, and tease out the unique approaches to learning, knowing, and doing in each
knowledge system, so that these unique approaches can be enabled. As a beginning, Dunne and
Dougherty (2006) delve into how industrial life scientists learn for drug discovery, and contrast
their mechanisms with those the literature says are used by engineers and academic scientists.
The literature suggests, for example, that technologists search for answers to defined problems
using measurable criteria, while academic life scientists (per Knorr Cetina) engage in blind
variation because the life system is unknown. Our findings indicate that industrial life scientists,
who must generate useful results like technologists, combine these strategies into a third one that
we label “searching for clues.” They proceed in an emergent, stepwise, but deliberate fashion to
feel out possibilities in a way that preserves the whole problem space but begins to zero in on
options. We develop other differences for integrating and sensemaking too, and infer from them
very different managerial approaches for enabling that learning.
Non-Decomposable? Another trap exacerbates the failure to see distinct systems of
knowing and doing in R&D – the presumption of decomposability. The history of the pharma
industry suggests that, following the logic of decomposition, people hived off different aspects of
the process and run them down until diminishing returns set in (e.g., rational drug design rather
than mucking about in natural products; micro-biology to zero in on cellular functions – both
took people away from the whole body). Diverse systems of knowing and doing seem to have
been added over the past 70 years in an ad hoc fashion, without much thinking about the whole
process, since each system emerged to provide a new set of answers (all in the “stick-it-in”
fashion of techno-hype). Now, most pharmaceuticals are reorganizing into multi-disciplinary
therapy teams to capture interdependencies since the life system cannot be decomposed. Each
drug must fit into the complex, massively redundant, utterly unpredictable life system without
triggering side effects, while the body compensates, diseases evolve, and individuals vary.
However, decomposability rather than integration is the hallmark of 20th century
technology management, where technology is “rationalized” and chunked up into separate
components and steps. Thus, work is specialized, simplified, and optimized. Academic life
science also works by fragmentation according to Knorr-Cetina (1999), where the objects of
study are detached from their natural environment and transformed for the lab. The challenge for
management academics is how to organize and manage one giant, holistic blob of
interdependent, interacting activities that takes up to sixteen years to play out for each drug.
Time is not decomposable either! What can we optimize? One possibility is to optimize both
the unique contributions of each knowledge system and their simultaneous, ongoing
intertwining. To that end, several more focused research problems must be attacked: identifying
the core knowledge systems involved (in each industry); figuring out each one‟s unique
contributions; figuring out how each enables and contributes to the others; and finally figuring
out how these systems can be integrated.
For bio-pharmaceuticals, I think the knowledge systems include academic sciences, the
industrial life and chemical sciences (perhaps two large groups but there may be subsets), the
technologies (also perhaps separated further into subsets), strategy, and process management.
The therapy areas such as pain, cancer, metabolism, or cardiovascular are also knowledge
systems. But they may function as boundary “spaces” that can help integrate the other systems,
and so can be examined from that perspective.
Regarding unique contributions, the sciences indicate where and how to search for drug
possibilities in the human body by providing understandings of disease processes and the
functioning of proteins, pathways, cells, and many other biological and chemical processes.
Most of the basic scientific research occurs in universities, government labs, and research
hospitals, but it must be intertwined with the industrial sciences in drug discovery to drive
toward finding a drug that modifies or cures a disease process. Academic insights are clues that
the pharmaceutical scientists combine with their own clues as they search for patterns, so there is
a real boundary between the two knowledge worlds. The scientific challenge is to continually
deepen and adapt explanations to match with empirical findings, so the explanations can absorb
more of what becomes known. The sciences are also essential for making sense of results of
experiments and of the outputs of the technology search engines. The sciences may differ in
their „epistemic communities‟ even within the bio-pharmaceutical work, so how the various
sciences themselves go together is a true challenge.
Technologies provide answers to the scientific questions. Technologies are scientific
instruments that construct the predictable conditions in which scientific explanations can match
the world, as Nightingale explains. In this role they collect and process (and in part determine)
the scientific data. For example, technologies produce images of how molecules fit into
receptors, models for chemical structures, and bio-informatics synthesizes characteristics of
successful drugs that might predict clinical success. Second, technologies “industrialize” the
science by mass production of biological and chemical materials and vast throughputs of tests
(e.g., to see what kinds of compounds may interact with what proteins, how compounds affect
other aspects of the life system, whether or not they might be cancerous). Third, technologies
provide drug delivery (injections, ways to convey a non-soluble compound through the gut) or
improve some aspect of the drug such as excretion or absorption. Fourth, technologies can
represent the human body (i.e., mice that are missing certain genes, bio-markers, genomics and
proteomics). Most technology outputs are inputs for sciences‟ clue searching, iterating into
packages, and sensemaking. Technology directly enables science by representing the life system
realistically and by bounding scientific clue searching, and may provide other unique inputs yet
Strategy is essential for guiding and shaping various searches, and somehow needs to
capture and make sense of the risks rather than simply work as a portfolio. The risks include
dealing with the unknown or managing what they do not know, dealing with the right questions
at the right time, dropping projects without eliminating major new possibilities, considering the
better domains of search and why (i.e., therapy area, disease pathways, targets and target types),
and what happens when certain questions are not addressed by certain times. The process
concerns moving projects along more effectively, but also recognizing the non-linearity of it all.
Regarding the connections, how to intertwine is a central problem. First, if we
understand the connections among the knowledge systems, these can be “fired off” like
lightening strikes so people can see the whole life system in that flash. However, intertwining
will always take enormous judgement and is inherently cerebral but collective, not only hands-on
“learning by doing.” But these complex judgements can be more systematically informed.
Knowledge would accumulate in each of the systems, but what knowledge and how needs to be
figured out. It would also accumulate in “platforms” that connect systems. The knowledge
integration is, I think, almost entirely social so material objects and data bases will play only a
supporting role. But everyone must do his or her own integration (so it is individual too).
Non-Scalable? A third trap of current models is the presumption of scalability, which
refers to standardizing products so parts are mass produced and the overall system can grow in
size or volume straightforwardly. An early observation from the pharmaceutical interview data
was that people were managing each compound almost independently from the others – when
asked about learning, one manager said that all the knowledge was drilled into “the compound.”
My first reaction was that this was wrong since in technology-based innovation, we have learned
that such micro management of products leads to unmanageable product proliferation. However,
if what they are doing is non-decomposable, then each compound has to be unique since it
interacts with the complex life system in unique ways, and there can be no simplifying
frameworks. As well, drug projects morph during the course of development, sometimes
radically (e.g., from treating multiple sclerosis to arthritis), so drug discovery is like starting to
build a car but ending up with roller skates.
The challenge is to figure out how to standardize without standards, or knowledge cannot
accumulate. Non-scalability encourages drug discovery people to look down into each
compound rather than up and out into their knowledge of the system. One approach is to shift
focus from the external frames (there are none) to the internal drivers of social action, which
perhaps can be leveraged across projects. One driver might be managing the questions being
asked, and trying to see what are the good ones to ask when and why, and what is the minimal
set of questions for different types of compounds at different stages. A related driver is that they
are learning to not fail rather than learning from failure, so how can they see what they do not
know better? And how can these ideas accumulate? And while problems cannot be surfaced
(they have multiple causes), perhaps problem setting can be surfaced and examined.
To conclude, academics have nailed the very important managerial principles of: 1) high
volume, high quality manufacturing; 2) trajectories and learning curves for technological
emergence; and 3) platforms and architectures underlying product and process innovation. But
these successes concern the special case of 20th century innovation that is based on technology
and engineering. People in bio-pharmaceuticals must figure out how to manage simultaneous
intertwining of disparate systems of knowing and doing for hundreds of projects spread over a
12-16 year cycle that itself ripples over time. So far they have little help from us “experts” in
innovation and knowledge management. If we as a community begin to really address their
problems, we will significantly advance our own models and our contribution to our students.
Dunne, Danielle and Deborah Dougherty (2006) learning for Innovation in Science-Based
Industries: The Case of Pharmaceutical Drug Discovery Rutgers University Working Paper
The Economist, June 18 2005, Prescription for change, a survey of pharmaceuticals (special insert); June
4 2005, From seed to harvest, p. 63.
Knorr-Cetina, K. 1999. Epistemic Cultures: How the Sciences Make Knowledge. Harvard University
Nightingale, P. 2004. Technological capabilities, invisible infrastructure and the un-social construction of
predictability: the overlooked fixed costs of useful research. Research Policy. 33 1259-1284.