The Method of Convergent Realism, Part II, Chris Santos-Lang

Step 3: Provision of Means by which New Discoveries will Force Retesting

Science, mathematics, and ethics each have a history in which some claims which were tested many lifetimes ago are now considered outdated. New background assumptions became available, and retesting revealed that the Earth is not the actual center of the universe, geometry is not necessarily Euclidean, and moral authority does not actually derive from gender or bloodline. If such retesting never happened, then science, mathematics, and ethics would converge on dogma, rather than on truth. One attempt to quantify this risk in science surveyed articles published in the top 100 psychology journals and found that only about 1.6% of the articles documented retests, yet retests by the original authors failed 10% of the time (Makel, Plucker, and Hegarty 2012). The implied convergence on 90% truth (or worse) is called the “replication crisis” because retesting is the implied solution … [please read below the rest of the article].

Image credit: NASA Goddard Space Flight Center via Flickr / Creative Commons

Editor’s Note:

Chris Santos-Lang’s “The Method of Convergent Realism” will be presented in two parts. Please find below Part II. Please also refer to Part I. The PDF of the entire article is linked in the Article Citation.

Article Citation:

Santos-Lang, Chris. 2022. “The Method of Convergent Realism.” Social Epistemology Review and Reply Collective 11 (1): 33-49.

🔹 The PDF of the article gives specific page numbers.

This supposed “replication crisis” highlights the fact that being subject to retesting is not the same as actually getting testested. While Step 2 focuses on the independence of retests, Step 3 focuses on making sure those retests actually happen. Critical retests may require new background assumptions which were not available in the lifetime of the original researchers, so how could researchers possibly be responsible for Step 3? It is through “responsibilities of citizenship” that researchers maintain the fields which will retest their reasoning as background assumptions change. Some of the acts of good citizenship these responsibilities demand of a researcher may seem unrelated to that researcher’s specific claims, but it would be unreasonable to expect to converge on truth if the responsibilities of citizenship are not met, and therefore would be unreasonable to make any claim without performing acts of good citizenship.

Specifically, this section argues that the following five acts of good citizenship are necessary to accomplish Step 3: transparency, conflict resolution, amendment process, expiration dating, and objectivity.

Act 1: Transparency

Transparency is about how well reasoning is explained. If the author of a claim is the only person who can understand its reasoning, then independent retesting becomes impossible, and articulation of background assumptions becomes useless. Even less extreme degrees of opaqueness can undermine the reasonableness of expecting convergence. For example, even a well-written explanation would be practically useless if lost in a fire or buried in “the Internet.” BREC-style procedures require publishing opinions and all related materials (with any sensitive information redacted) to the Open Science Framework (OSF), where they are encrypted and backed-up twice a day to multiple locations. The OSF is backed by a $250,000 preservation fund which is expected to ensure free access to its data on the World Wide Web for at least fifty years. All of these details are important. To be transparent includes making one’s publication intelligible, accessible, discoverable, citable, and permanent (but revisable). Researchers may someday be able to achieve all of these aspects of transparency through direct communication with something like Watson, but, regardless of how transparency is achieved, it clearly would be unreasonable to expect Watson to converge on truths it never encounters.

BREC may have been the first IRB to publish its opinions so transparently. Here’s why all research ethics committees should do the same: Suppose a researcher submitted a research plan for review, and the research ethics committee discovered an ethical problem with it, but kept that discovery between the researcher and the committee. While the committee prevented that researcher from causing harm, other researchers may stumble onto the same idea, find no record of your discovery, implement the unethical plan, and end-up inflicting the harm. The committee could have prevented more harm—maybe even saved lives—if it published its opinions more transparently.

One theme to notice in Step 3 is that acts of good citizenship often support research not “owned” by the actor. For example, transparency isn’t just the act of the research ethicist who posts information to the Internet when they discover a reason not to conduct an experiment. Transparency is also the act of the truck drivers who deliver the materials to generate the electricity that powers the servers which archive and index that post a century after the research ethicist passed away. Transparency is becoming a group effort and a struggle that will outlast any of us individually.

Act 2: Conflict Resolution

The second act of good citizenship is conflict resolution. As an example, consider the conflict Marc Edwards resolved between Lee Anne Walters and the city engineers of Flint, MI, over the quality of its water supply. Lee Anne was a resident of Flint, tested her water, and found it unsafe. She was not a professional scientist, but she asked others to conduct independent tests. While many other non-professionals replicated her results (i.e. Step 2 occurred many times), many professional scientists refused to get involved, and the Flint engineers refused to fix the water supply. The conflict was resolved when Marc Edwards, a professional scientist from Virginia, repeated Lee Anne’s test and confirmed her conclusion. The Flint engineers then fixed the water supply to the satisfaction of both Lee Anne and Marc.

It is worth noting that the water supply was not fixed to the satisfaction of Scott Smith, another non-professional who tested the water, and that Scott and Marc therefore found themselves in a justifiably passionate dispute over whether Flint residents should be told not to bathe. Hygiene is important for public health, so whomever was wrong in this debate was endangering Flint residents. Marc became threatened with life-altering legal action, until Scott conceded that bathing would be safe. Further discussion about Scott will be deferred to the section about objectivity, but he is mentioned here to demonstrate that it would be unreasonable to expect the city engineers to trust every report from people who think they are conducting science and that it would be unreasonable to expect all professionals to subject themselves to the same danger Marc braved. Fortunately, many professional scientists are just as qualified to do what Marc did–the Flint situation had little to do with Marc’s own research. But if none of them perform the act of conflict resolution, then Watson would be as stuck as the Flint engineers, unable to sort through the conflicting claims.

The BREC procedures do not demonstrate that it is possible to reliably generate peacemakers, but one of the ways in which the BREC procedures improve over traditional procedures is to provide a mechanism. In contrast, each traditional research ethics committee has a jurisdiction such as a school, hospital, or laboratory. If two such committees reach contrary conclusions about the same research plan, then the research will be allowed in one jurisdiction but not in the other. Thus, a research plan with global risks (e.g. risk of creating a pandemic) could be forbidden in ninety-nine jurisdictions but permitted in the hundredth. Thus, the procedures of traditional research ethics committees converge on all research getting permitted somewhere eventually. In contrast, the BREC procedures state that BREC opinions can be suspended or terminated by “any credible committee,” and provide simple instructions to form additional committees as needed to resolve conflicts. The concept of “credible committee” may bring the same frustration that the concept of “professional scientist” brought to Flint, but the BREC procedures at least allow a mechanism by which a risk of global harm could be blocked globally.

Act 3: Amendment Process

The third act of good citizenship, maintenance of an amendment process, involves making it feasible to reverse one’s claims. It can take effort to avoid reaching a point of no return. For example, environmental scientists maintain seed banks, isolate nature preserves, and protect endangered species so that any damage can be undone if accepted reasoning is later found invalid. The classic example in which ethics suffer for lack of an amendment process is when the government of a state, religion, or business cannot be sufficiently amended to avoid corruption, so that state, religion, or business finds itself at odds with truth.

A common dystopian vision for AI involves someone corrupting it, permanently bending the AI to its own agenda, thus preventing the AI from ever converging on certain truths. Typical strategies to avoid such exploitation boil down to decentralizing networks, as in blockchain or the World-Wide Web, such that coordination is achieved more through open standards than through any central database that a corrupting entity could control. BREC is architected this way. It has procedures for amending opinions and for amending procedures, but it takes a further step analogous to open source software: Its infrastructure can be “forked” with a single click.

In contrast, despite being digital, modern journals do not automatically amend articles when their cited sources are redacted. Furthermore, some industries do not quickly and appropriately amend themselves to match discoveries about climate change and ecosystem change. Some governments do not quickly and appropriately amend themselves to match discoveries about threats of contagion and the efficacy of masks. Some religions do not quickly and appropriately amend themselves to match discoveries about social polarization. Institutions that are not well-maintained will become corrupt, resist truth, and battle against anyone who does converge on the truths they resist, so no one can reasonably expect to converge on truth without maintaining (or limiting) institutions.

Act 4: Expiration Dates

The fourth act of good citizenship is to record the dates of tests and retests. Test dates imply expiration dates: If the most recent retest occurred long ago, then one must ask whether better background assumptions have since become available. That is a good reason for researchers to date their publications, which is what most researchers do, but many researchers neglect to publish retest dates. For example, a given mathematical claim may be retested each year by hundreds of college-level mathematics classes (i.e. proving a theorem on a chalkboard), but its most recently published retest may be decades old.

If expiration dates were recorded, then AI like Watson could use those records to flag information as “expired” and to tell researchers (and funders) what needs to be retested. But, if expiration dates are not recorded, then Watson might forever hold a false claim that never gets retested.

No innovation is necessary to demonstrate the feasibility of tracking expiration dates. This is already standard in research ethics. Following these standards, the BREC procedures require all opinions to be renewed at least annually, and requires that the implied expiration dates be published on all consent forms. Such fast expiration may, in fact, be overkill for most research plans.

Act 5: Objectivity

The fifth act of good citizenship is objectivity. Some people may think science and mathematics are objective by definition, but the relevant kind of objectivity is the range of people who can test a claim. For example, claims about whether a work of art invokes a particular feeling in a particular person are considered subjective because only that person can test those claims. Yet mathematical reasoning would likewise lack objectivity if only a rare supercomputer could test it, and science that relies on rare spaceships, supercolliders, or giant datasets may be interesting, but is relatively non-objective. We perform the fifth act of good citizenship by widening the range of people who can access equipment and facilities required to retest claims. It starts simply with making spaceships and other resources less expensive, but then progresses towards open-sourcing software and hardware and towards establishing community labs.

The section on conflict resolution deferred further discussion of Scott Smith until now. Scott initially disagreed with Marc Edwards about the quality of the water in Flint, Michigan. Marc was not allowed to independently retest Scott’s claim because Scott’s test of the water quality used new proprietary technology he had developed. That sounds fishy: It is possible that the new test doesn’t work. However, it is also possible for someone to develop a better test, and the typical way to secure compensation for such an invention is to restrict others’ access to it (i.e. by sacrificing objectivity). This is an example of how lack of objectivity can prevent an AI like Watson from converging on truth by blocking the AI from telling which of the two possibilities is the actual truth.

Objectivity isn’t just about increasing access to new technologies. It’s also about keeping costs low for technologies that are already accessible. For example, researchers themselves are a technology that seem plentiful today, but they could become rare without ongoing parenting, educating, growing of food, and maintaining of shelters, medical services, security and financing. If researchers became rare, then retesting would become expensive, and whatever errors exist in the data used by Watson and other AI might never get corrected. Thus, perhaps the most common act of objectivity is to raise the next generation. The people who perform this act vastly outnumber the people who call themselves “researchers”, but play an equally essential role in the method of convergent realism.

The BREC procedures create the equivalent of community labs for research ethics. Most IRBs rely on some volunteers, but the BREC procedures eliminate the non-volunteer labor by leveraging the Open Science Framework, so a remarkably wide range of people can afford to assemble an equivalent committee and retest any BREC reasoning. In contrast, the Association for the Accreditation of Human Research Protection Programs reported that the median IRB budget in 2016 was about $800,000 annually or $400 per experiment (2017). This practically limits retesting to wealthy or institutionally-supported researchers. The reasoning of centralized courts such as a federal agency, Congress, or Supreme Court is even less retestable, since it would be even more expensive to assemble an equally-qualified group of people. People who wish to converge on truth—rather than on something distorted by politics—thus have good reason to treat centralized courts as a last resort.

Units of Convergence

Calling transparency, conflict resolution, amendment process, expiration, and objectivity “acts of good citizenship” raises the question, “Citizenship in what?” The answer to this question might be labeled the “unit of convergence” because it is the entity that is to converge on truth. The method of convergent realism requires transparency and conflict resolution within that unit, but does not require transparency and conflict resolution beyond it. For example, if the unit were humanity, then the method would not require humans to be transparent to whatever aliens (if any) exist on the other side of the universe. Likewise, it would not require us to resolve conflicts with aliens, maintain alien institutions, or make retests affordable to aliens.

Analogously, a unit of convergence could be smaller than the whole of humanity. Step 1 merely requires that it include parts that can articulate reasoning. Meanwhile, Step 2 merely requires that its parts be sufficiently independent to mitigate each other’s biases and Step 3 merely requires that each unit of convergence be able to retest all claims indefinitely. Thus, if separate nations or companies could coexist indefinitely, then it could be reasonable for each to expect to converge on truth independently. Likewise, if an AI or individual human could sustain itself indefinitely and contain enough diversity to mitigate all biases, then it might reasonably expect to converge on truth all by itself. However, it seems unlikely that the method of convergent realism would settle on such individualism.

The pressure against individualism holds an analogy to the dynamics hypothesized in integrated information theory. According to this theory, consciousness comes in differently sized grains, such that consciousness of warmth might be of smaller grain than consciousness of a book because it requires more mental resources to recognize something as being a book (Tononi et al. 2016). The evidence to date concurs that removing pieces from a person’s brain does not eliminate their consciousness, but does reduce the number of things of which that person can be conscious. Much as different grains of consciousness may require different amounts of mental resource, different claims are retested at different costs (i.e. some require a supercollider, larger sample size, etc.). Thus, assuming maximum efficiency, a larger unit of conversion can reasonably expect to converge on more truths simply because more retests fall within what it can afford.

The “maximum efficiency” caveat above is significant. Each citizen almost constantly impacts the transparency, conflict resolution, amendment process, expiration, or objectivity of their unit of convergence. A citizen might not participate in every step of the method of convergent realism, but there are steps like “raising the next generation” (a part of objectivity) where fields overlap, where we cannot avoid having impact, and where one’s only choice is how positive or negative one’s impact will be. Some potential citizens might not even have much choice about that–they might not be able to contribute positively. For example, humanity as a whole includes some individuals who are too developmentally handicapped to learn language. Even tools like Watson cannot empower such individuals to overcome their handicaps, so they become excluded from the unit of convergence, at least regarding some claims. They are treated like children for their entire lives, never receiving complete transparency, never having their conflicts taken seriously, and never really being empowered to test certain claims for themselves. Regarding the most “advanced” claims of science, mathematics, and ethics, it might be argued that is how experts treat even the average human being.

Some elitism is unavoidable; we would go bankrupt trying to empower every member of every species to retest every claim we currently hold as true. On the other hand, if soybeans nourish independent testers, that might qualify soybeans as citizens of our unit of convergence, even though their contributions to convergence would be highly specialized. Soybeans and humans are not “equal” citizens in the sense of having equal opportunity to contribute in each role, yet soybeans may contribute more than certain humans, even more than certain brilliant humans. That’s because a unit can be forced to exclude a potential citizen simply because that potential citizen’s arrogance blocks that potential citizen from being transparent with, resolving conflicts with, or making retests affordable to other potential citizens.  For example, imagine a nation that withholds strategic scientific discoveries or resources from the rest of the world. Such arrogance may force a unit to choose between potential citizens, thus escalating a conflict between individuals (or between individual nations)  into a conflict between individualism and communalism (or nationalism and globalism). Conflict over what roles may be afforded to AI could likewise escalate into conflict between individualism (or speciesism) and communalism. Communalism survives such conflicts by undermining objectivity, increasing the dependency of individuals on communities. Yet we cannot reasonably expect to converge on truth when converging on subjectivity, so arrogance seems incompatible with reliable convergence on truth.

Conclusion and Potential Responses

This essay articulates a perspective that might not be shared by all readers. Some readers may doubt the existence of natural laws or facts of science, mathematics, or ethics. Others may doubt the existence of any reliable method we could follow to discern those laws or facts. Finally, some may claim that there is a way to accomplish such convergence without following all of the steps mentioned here. Please share any such alternative method on the PubPeer page assigned to this essay so  all readers can find them.

Yet other readers may consider the ideas in this essay trivial. “Of course we would have no reliable reason to expect Watson to converge on truth if the company feeding it new information will not last,” they say, “That’s why the Watson engineers chose to work at IBM, a company expected to last a long time.” Yet even IBM might not last forever, so there are arguments to be made that Watson should be engineered to be transferable to an open source community, and to help humanity address existential threats such as climate change and nuclear stockpiling. This essay provides formalization to make such arguments rigorous. Moreover, comparing BREC procedures to those of traditional research ethics committees demonstrates that the formalization in this essay can be applied to yield practical innovation.

Other readers may share the perspective articulated here and appreciate its practical steps. They may believe it has meaningfully elaborated beyond previous attempts to describe a method of convergent realism. Ideally, they will share this elaboration among a unit of convergence, and elaborate even further. [Please refer to Part I of the article.]

Author Information:

Chris,, is the author of several patents and former co-chair of the Ethics Working Group of the Citizen Science Association.


Association for the Accreditation of Human Research Protection Programs, Inc. 2018. “2017 Metrics on Human Research Protection Program Performance for Academic Institutions.” Washington, DC. 2017%20Academics%20Metrics.pdf.

Belleville Research Ethics Committee Procedures. 2017. IRB 11228.

Committee Opinion about Replication of Merolla et al. 2017. IRB 11228.

Dewey, John. 1910. How We Think. Boston, MA: DC Heath.

Gerstein, Dean R. 2016. Correspondence. ClaremontLetter.pdf (Version: 1)

Hardin, Clyde L. and Alexander Rosenberg. 1982. “In Defense of Convergent Realism.” Philosophy of Science 49 (4): 604-615.

Hoffman, Sharona and Jessica Wilen Berg. 2005. “The Suitability of IRB Liability.” Case Legal Studies Research Paper No. 05-4.

Kitcher, Philip. 1990. “The Division of Cognitive Labor.” The Journal of Philosophy 87 (1): 5-22.

Klitzman, Robert. 2012. “Institutional Review Board Community Members: Who Are They, What Do They Do, and Whom Do They Represent? “Academic Medicine: Journal of the Association of American Medical Colleges 87 (7): 975–981.

Lintott, Chris, Kevin Schawinski, Steven Bamford, Anze Slosar, Kate Land, Daniel Thomas, Edd Edmondson, Karen Masters, Robert Nichol, Jordan Raddick, Alex Szalay, Dan Andreescu, Phil Murray, Jan Vandenberg. 2010. “Galaxy Zoo 1: Data Release Of Morphological Classifications for Nearly 900,000 Galaxies.” Monthly Notices of the Royal Astronomical Society  410 (1): 166-178.

Laudan, Larry. 1981. “A Confutation of Convergent Realism.” Philosophy of Science 48: 19-48.

Longino, Helen E. 1990. Science as Social Knowledge: Values and Objectivity in Scientific Tradition and Change. Princeton: Princeton University Press.

Makel, Matthew C., Jonathan A. Plucker, and Boyd Hegarty. 2012. “Replications in Psychology Research: How Often do They Really Occur?” Perspectives on Psychological Science 7 (6): 537-542.

Popper, Karl. R. 1963. Conjectures and Refutations. London: Routledge.

Post, H. R. 1971. “Correspondence, Invariance and Heuristics: In Praise of Conservative Induction.” Studies in the History and Philosophy of Science 2: 213-255.

Putnam, Hilary. 1982. “Three Kinds of Scientific Realism.” The Philosophical Quarterly (1950-) 32 (128): 195-200.

Tononi, Giulio, Melanie Boly, Marcello Massimini, and Christof Koch. 2016. “Integrated information theory: from consciousness to its physical substrate.” Nature Reviews Neuroscience 17 (7): 450-461.

US Department of Health and Human Services. 2017. “Final Rule Enhances Protections for Research Participants, Modernizes Oversight System.” HHS Press Office.

US Department of Health and Human Services. 1998. “Institutional Review Boards: A Time for Reform.” Office of Inspector General Publication OEI-01-97-00193. June G. Brown, Inspector General. Washington, DC.

Categories: Articles

Tags: , , , , , , , , ,

Leave a Reply