# Q- What are the most efficient routes to useful cross-boundary knowledge

Last updated May 21, 2022

Authored by:: P- Joel Chan

## # Synthesis

• Hypothesis: we have far better evidence of people being the initial / most important first entry points into ideas across disciplinary (and other) boundaries, compared to search or recommendation that is separated from a social interaction of some kind. This is a variant of C- In interdisciplinary domains, network-based foraging is more powerful than search-based foraging
• We have lots of evidence of people being an effective (and in many cases the most preferred) channel to cross-boundary information
• In describing how scientists at an interdisciplinary center imported information from outside their specialty, @palmerWorkBoundariesScience2001 identified a range of practices, from talking to personal contacts, reading literature, and engaging in more formal learning activities. A strong theme was the special efficiency and effectiveness of talking to people to access and filter cross-boundary information with maximal signal-to-noise ratio.
• For example, Palmer writes, “no other type of information interaction is likely to be as efficient and profitable as interactive discussions with colleagues” (p. 33-34)
• She also summarized that scientists at the interdisciplinary research center imported information across disciplinary boundaries predominantly via personal networks (pp. 31, 32-35). For example, A device physicist at the center said that the vast majority of their information came from keeping on contact with people and going to conferences
• There is also research on sources of innovation in innovation contests and firm-/organizational-level innovation that suggests an outsider innovation phenomenon: breakthrough innovations are more likely to come from people who are technically marginal/distant from the problem domain
• For example, @jeppesenMarginalityProblemsolvingEffectiveness2010 found that Solvers who chose to submit their solutions to the 2nd stage of an innovation contest and self-rated themselves as more technically distant from the problem domain were slightly more likely to submit winning solutions for Innocentive science problem solving contests
• Additional indirect evidence comes from the literature on collocation and collaboration, which so far we can see generally seems to find facilitative effects on knowledge transfer and collaboration from induced face-to-face contact
• An entry point is @cataliniMicrogeographyDirectionInventive2017, who found substantial changes in the likelihood of collaboration across research labs as a function of collocation, particularly for labs that were “further apart” in research areas.
• This work suggests / is consistent with the idea that other channels of search across boundaries are less efficient than face-to-face (presumably “small touch”, bc collocation) contact with people from other units (i.e., across boundaries). the logic is that if the search costs were the same / comparable, then those ideas would have initiation probability that is similar to collocation, in the absence of collocation.
• Similarly, @boudreauFieldExperimentSearch2017 found that The likelihood that any given pair of scientists who were assigned to the same breakout room to discuss their work would collaborate on a pilot grant proposal was estimated to be substantially (~75%) higher ($$p(collaborate) ~= .28$$) compared to scientist pairs who were assigned to different rooms ($$p(collaborate) ~= .16$$). This estimated effect was statistically significant at p < .10.
• Note, however: this effect seems to be driven by facilitating scientist pairs who are from the same clinical area to collaborate, which suggests there might still be important inefficiencies to accessing and using cross-boundary knowledge, even when you get initial access to it through a person.
• We lack evidence that literature discovery tools that don’t involve tight integration with at least one person can actually help people cross knowledge boundaries
• Why might people be especially efficient access points to cross-boundary information?
• Note that there are still some inefficiencies for the people route
• For example, @laneEngineeringSerendipityWhen reported that scientist pairs who were randomly assigned to talk to each other (and did so, face-to-face) at an interdisciplinary medical imaging symposium used about 6 new MeSH keywords from their partners in future publications when the pair had moderate overlap in intellectual interests, compared to low or high overlap (for which there were no clear effects)
• It’s notable that the observed positive effects here were for estimated less than 20% of the scientist pairs!
• And in some cases (e.g., when field similarity was high), there were dampening effects on co-authorship in the 6 years following the symposium.
• #appraisal - these results seem a bit fragile to me: they rest on interaction terms in relatively complex OLS models.
• Caveats aside, this result suggests that the likelihood that a randomly paired interaction, even within a relatively shared community of interest, and estimating effects on a relatively “lightweight” outcome (i.e., evidence of knowledge transfer, instead of more costly endpoints like copublication), is probably less than a coin flip!
• These results raises the question: Q- What is the range of expected “benefits” of engineered interactions with potential “informants” across knowledge boundaries

## # Sources

Made by Rob Haisfield, Joel Chan, Brendan Langen using Quartz, © 2022