Monday, March 16, 2015

Access to Error

Guest Blogger

Michal Shur-Ofry

For the Innovation Law Beyond IP 2 conference, March 28-29 at Yale Law School

What are your immediate associations to the word "innovation"? The responses I received while running this question through friends and acquaintances ranged from the general ("progress", "future", "modernization", "technology", "patents", "curiosity", "thinking outside the box", "fresh", and "intriguing") to the specific ("artificial intelligence", "spaceships", "cherry tomatoes", "Einstein", "Da Vinci" and "Apple"). No one, however, mentioned errors, failures or negative findings.

This is hardly surprising—we tend to associate innovation with success (or other positive things) and errors with defeat. But errors and innovation are actually tightly linked. My purpose in this project is to focus on errors as drivers of innovation. I argue that the current incentive structure in our innovation ecosystem, both within and beyond IP, does not provide sufficient incentives for the dissemination of errors and other negative information. I further hope to start a conversation about access-to-error as an important (and largely overlooked) goal for innovation policy.

First, what do errors have to do with innovation? To clarify, I use "errors" here in a very broad sense, that includes mistakes, failures, falsifications, blind-alleys, negative findings and additional types of negative information. The main answer is almost obvious, and was suggested by philosophers of science long ago: errors provide us with important negative knowledge, the knowledge of what doesn't work, which brings us closer to understanding what does. In the words of Karl Popper: "we learn from our mistakes."

But there is more to it: errors are especially important for triggering paradigm shifts---a particular type of innovation that opens up new fields of research and can completely change scientific domains. Thomas Kuhn in his influential work about scientific revolutions recognized that paradigm shifts are often preceded by detecting mistakes and inconsistencies under existing paradigms. Indeed, actual shifts in physics, life-sciences and even behavioral economics provide ample examples for the power of errors to push innovation beyond the state of the art. More recent research in the field of complexity highlights another angle: due to the networked and inter-dependent nature of many innovation ecosystems small errors that accumulate undetected can eventually cause large-scale catastrophes---famous failures of aircraft, buildings and nuclear plants provide powerful examples. The detection of errors in complex innovation ecosystems may therefore be especially significant. Finally, although my focus is on the scientific and technological domains, errors and mistakes are also drivers of artistic creativity.

Of course, the potential of errors to promote innovation depends very much on their disclosure and dissemination, or in other words, on access-to-error. But a quick look at the main legal and social institutions that promote the diffusion of knowledge reveals that these institutions hardly incentivize the diffusion of errors. First, intellectual property, a primary legal mechanism for incentivizing the dissemination of knowledge goods is largely dysfunctional when it comes to negative information. This is because IP rights are based on exclusion (i.e., on the ability to control the use of the protected subject matter and prevent unauthorized uses), while negative information is generally difficult to exclude—as articulated by Amy Kapczynski and Talha Syed, it is located at the lower end of the continuum of excludability. It is often impossible to trace those people who actually used certain negative information, and distinguish them from other persons who avoided the wrong path for various other reasons.

What is more surprising is that the major alternatives to IP---the scientific establishment with its reputational rewards, and state-supported grant and funding schemes—also fail to incentivize the disclosure and dissemination of negative information. In fact, scientists face substantial difficulties when they try to publish negative findings. Accumulating evidence from different fields suggests that in many high impact journals "negative results are not accepted", and often, negative findings never see daylight (see, for example, recent evidence by Daniele Fanelli that "negative results are disappearing from most disciplines and countries"). And even when published, negative information attracts less citations and less public attention than positive data.

Similarly, many state-supported grants and funds strongly prefer projects that seek to attain positive findings over projects which focus on falsification or replications. Leaders of prominent funds quoted in the Economist explain that the latter "in all likelihood would be turned down" (one encouraging exception may be a forthcoming change in NIH policies, as announced by heads of NIH, Francis Collins and Lawrence Tabak, in a recent comment in Nature).

This current incentive structure, combined with common cultural perceptions of error as shameful and humiliating, produces a well-documented "file-drawer effect": positive information is disclosed ad disseminated while negative information is shelved into a huge metaphorical file-drawer. And the social costs are significant: distortions of the "big picture" of science and technology; troubling evidence that much of what we think of as proven scientific knowledge is actually wrong (John Ioannidis even maintains that "most published research findings are false"); innovators who are steered away from potentially ground-breaking projects toward "safer" and more conservative enterprises, and valuable R&D resources that are wastefully poured into blind alleys. In some cases, the difficulties to disseminate and access negative knowledge may even risk lives: flawed scientific hypotheses "live on" and create health hazards; drugs with dubious safety may find their way to the market, and undetected errors in complex systems accumulate "under the surface" until abruptly causing disastrous outcomes.

How can innovation law and policy facilitate access to error? This project does not provide a comprehensive solution, but sketches in rather broad strokes three principal directions for policy intervention. First, I propose a few adjustments to IP doctrine (mainly, to patent law's disclosure requirements and to IP's exceptions and limitations) whose essential purpose is to lower access-barriers and allow the exposure of errors in IP-protected technologies without risking IP infringement.

Second, I explore the option of direct, top-down, regulation mandating the disclosure of negative information. While this type of state intervention has significant drawbacks as a general solution, it may be suitable in certain settings, especially where disclosure pertains to safety and efficacy and regulatory schemes are already in place (an amendment to FDA regulation that instructs the disclosure of "adverse events" that are part of clinical trial results is one such example). More generally, highlighting the links between error and innovation can alert regulators to the significance of negative information, and help them calibrate their regulatory efforts accordingly.

Finally, the most promising direction for promoting the diffusion of errors may be locating access-to-error within the larger paradigm of access-to-knowledge. Building on the extensive scholarly work of the recent decades I suggest that the norms of sharing cultivated by the access-to-knowledge movement can serve as a useful model for commons-based schemes designed to improve access-to-error. Emerging enterprises in various domains (such as the Journal of Negative Results in Biomedicine or the Reproducibility Initiative) indicate that this could be a productive path.

Such schemes are not susceptible to top-down state regulation and their success largely depends on the emergence of sociocultural patterns. But the state can still play an important role here, by adopting a variety of measures to facilitate commons-based projects and nudge bottom-up initiatives of access-to-error. Concrete examples for such measures which I explore in the paper include:
  1. initiating and facilitating the formation of errors-repositories
  2. calibrating state funding schemes to support the access-to-errors paradigm
  3. adopting and supporting voluntary "near-miss" schemes for reporting errors
  4. promoting a broader educational agenda that embraces error rather than conceals it.
It is time that access to error becomes a prominent item on the innovation policy agenda. This project, I hope, is a step in this direction.

Michal Shur-Ofry is Senior Lecturer, Hebrew University of Jerusalem, Faculty of Law. She can be reached at michalshur at


Older Posts
Newer Posts