Digitising reflective equilibrium

It is when we consider what balancing all these considerations would practically entail that the trouble starts. Theorists should be charting all the possible interconnections between an extensive set of judgements, a further set of ever-developing principles, and numerous background theories, constantly revisiting and revising all of these elements as they go. They will presumably also need to reliably record what combinations they have already tried, why different elements were accepted or rejected, as well as which combinations remain up for evaluation. If that is not already enough, this complexity will only be compounded by the need to repeatedly weigh elements against each other in a cyclical manner, as any adjustments in the system may always have knock-on effects elsewhere.

As anyone who has ever tried will attest, it therefore takes an awful lot of work to reach a stable RE. That is, if reaching one is achievable at all—Rawls (2005, p. 97) famously thought “the struggle for reflective equilibrium continues indefinitely”, as there would always be more content to integrate down the line. Yet despite how “demanding” or potentially even “unachievable” finding an RE is, it is nevertheless considered the ideal that theorists should be striving for (Knight, 2017, pp. 49–50). You quickly begin to wonder if this is why “few real applications of the method are presented in the literature” (Doorn & Taebi, 2018, p. 508). Time-poor, modern academics simply cannot commit years of their lives to crafting the sprawling, complex equilibrium that could underpin their own A Theory of Justice. Indeed, I suspect few theorists ever actually follow the RE process to completion, which perhaps explains why even fewer have thoroughly, openly, and transparently published their attempts to do so in print³.

One way to ease the burden is of course through the employment of external aids. Perhaps, in neatly circumscribed situations, some exceptionally talented individuals may be capable of constructing narrow reflective equilibria within their own heads (it is certainly not possible for me). But I doubt anyone could reach a wide RE without technical assistance. At the very least, pens and paper, whiteboards, spreadsheets, or word processors are going to be required to record (a) which judgements have been selected or rejected, (b) which principles have been tried, tweaked or tossed aside, (c) how background theories might speak for or against the developing equilibrium, and (d) how all these considerations will interact. Evidently, tracking all these complex interconnections with traditional tools will be exceedingly unwieldy and time consuming.

The problem is that academic philosophical writing is decidedly linear. Our arguments generally unfold via continuous narrative threads, so the writing tools and approaches we are taught to employ are geared towards this linearity. But serialised sentences and paragraphs—recorded in reams of notes and lengthy Word documents—are poorly suited to capturing an ever-evolving web of interconnected judgements and principles (Fig. 1). There is no one starting point to an RE, with many possible routes into and through the mutually supportive web available. Even analogue tools supporting a more freeform approach cannot effectively handle the process’s sheer scale, with all its potential for rewrites and refactoring. The task is vast, multi-dimensional, and recursive, with each element usually offering holistic support to many of its neighbours. In short, there is a real practical challenge here for the theorist attempting to reach an RE with the help of traditional tools, rendering them ineffectual for realising the method.

To be clear: I am not suggesting to do away with the essay or monograph. Translating the circuitous back and forth of RE into linear form is certainly possible and, later in the process, even highly desirable. But it is important to recognise that it is far easier and more appropriate to do this towards the end of a project, when and if a coherent equilibrium that can be neatly summarised has been found. The theorist’s task then shifts: how can they best convey the most important parts of the RE they have constructed to their readers? Before then, though, it makes far more sense to work non-linearly, in a way that reflects the distributed structure of the RE process itself. Thankfully, digital tools supporting such an approach already exist. Although they were not developed for RE, they nevertheless offer features that can be readily adapted for its purposes.

Whilst mind-mapping programs might immediately spring to mind when faced with a non-linear problem space, I will instead focus on a subset of note-taking applications designed for ‘personal knowledge management’. My chief recommendations are Zettlr (Erz, 2023), Logseq (Numerous Authors, 2023), and Obsidian (Li & Xu, 2023)⁴. These ‘zettelkasten’ applications, which draw inspiration from the slipbox of notes maintained by prolific German sociologist Niklas Luhmann, possess toolsets that make them far more powerful than mind-mapping programs. As both are also free, enjoy committed online followings and operate on what are essentially plain-text ‘markdown’ files, they represent relatively safe and accessible options for modern academics. Online support networks are extensive and any basic text editor can open the files they create, so research data should remain accessible even if these particular programs are one day depreciated and replaced.

Repurposed for our ends, zettelkasten programs can empower theorists to build a database of notes that connect together in ways not supported by traditional note taking solutions, readily mimicking the structure of an RE. Of course, they include all the basic writing, editing, and formatting functions you would expect. In addition, both also support relatively unique features which make them particularly well-suited to realising the RE process. These are: inline bibliographic citations⁵, bidirectional hypertext links, hashtags, and graph visualisations (Fig. 2). Whilst only the first two features are perhaps essential for constructing what I will henceforth call digital reflective equilibria (DRE), the latter two result in significant quality of life improvements for theorists and their readers. Altogether, I will argue they push the utility of DRE far beyond their analogue ancestors.

The second feature is more exotic—and some version of this functionality is utterly essential for realising DRE. Bidirectional links allow theorists to mimic and map RE’s non-linear reasoning structure, capturing its constitutive interconnections digitally⁶. In zettelkasten applications, wrapping text in a double set of square brackets, like [[this]], converts it to an internal hyperlink the connects to another note in your collection, much like the links between Wikipedia pages. Clicking the link leads you to the other note, whilst also placing a ‘backlink’ which relates the target back to the origin. In effect, linking thus draws a virtual line between two notes, like string on a corkboard. Properly utilised, this functionality allows theorists to make explicit and track the connections between all the different elements of the RE process.

Let me provide an example. In the chain of notes a - b - c, note b could justify the link between a and c. In an RE referring to the trolley problem, for instance, it might read ‘killing is different to letting die’, and thereby connect the principle ‘killing is always wrong’ in note a with a judgement in note c that ‘diverting the trolley is unacceptable’. Such relational associations are central to the RE method, and so links support theorists to build up webs of interrelations between judgements, principles, and theories to make their interactions more easily understandable. Importantly, this is also achieved without requiring linearity—each note can have unlimited links, allowing the ideas it contains to connect to and play a part in justifying any other idea in the coherent web as required.

The other way I imagine tagging being used is to identify each element as either a #judgement, #principle, or #backgroundtheory when pursuing a wide RE. I should note, however, that although my earlier sketch perhaps suggested hard and fast distinctions between the various levels of abstraction in the RE process, these distinctions are only analytical. Theorists do not necessarily need to delineate their beliefs into neat silos before embarking upon the adjustment process—“there is no problem with the boundaries between [elements] being fuzzy or overlapping” (Knight, 2017, p. 53). Indeed, part of the RE process involves refining and revising initial commitments, even moving ideas between levels of abstraction as we promote judgements to principles or vice versa. Thankfully, as notes can have multiple tags and be altered at any time, tagging flexibly supports this ambiguity.

The final feature is graph visualisation. Many zettelkasten programs can generate a visual representation of the theorist’s web of notes, rendering any links as connections between nodes on a graph (Fig. 3). Unlike with more traditional methods of RE construction, this functionality literally allows the theorist to see, explore, and trace the connections between their ideas. Most mundanely, this can help them understand the overall ‘shape’ of their arguments, locating major branches and clusters. Perhaps more usefully, though, theorists would also be able to more readily identify either highly connected or completely unconnected notes, which will each require attention for different reasons. Notes with many connections, for instance, will enjoy greater prima facie support from numerous neighbours, pointing to their possible nodal importance within the developing web of coherent ideas. Conversely, unconnected or orphaned ideas will need to be better linked to others, or else outright rejected due to a lack of support. Identifying both will be made far easier by digitisation, thereby aiding the construction of well-supported, coherent, and holistic arguments.

Taken together, these features form the technical foundations needed for realising DRE. They will empower theorists to more thoroughly map and manipulate their ever-developing positions, with digital tools supporting greater systematicity, flexibility, and visualisability than alternatives. Much of this utility comes from being able to easily cite ideas, categorise as well as manage different elements, and link together a distributed body of notes to build up a coherent and navigable whole. By contrast, marshalling similar levels of oversight and control using word processors, whiteboards, or pens and paper would be exceedingly difficult, if not entirely unachievable. For these reasons, I suspect that the theorist working digitally is far more likely to be able to actually perform the intellectually demanding and practically extensive task of constructing an RE—particularly with regards to the substantial requirements of wide RE.

Whilst many of the principal benefits of DRE are thus essentially practical, digitisation also brings other advantages. For instance, Christoph Baumberger and Georg Brun (2021, p. 7935) have argued that one of “the most pressing internal challenges for defenders of reflective equilibrium” is the need to specify a more exact characterisation of the process of equilibrating. Although they submit that much of this will be project-dependent, Baumberger and Brun emphasise that, whatever the project, reconstructing the process of reaching an actual RE not only contributes indirectly to its justification but is also the only way for the theorist to develop “a precise configuration” of the rules for adjustment and epistemic goals that drove their theorising in the first place (Baumberger & Brun, 2021, p. 7937). In other words, it is only once a candidate RE has been found that we can retrospectively discern the most important drivers of the equilibrating process.

Working with DRE will make this evaluative task far easier. In fact, theorists that take full advantage of tagging and linking will find that much of their reconstruction work is already done, as these features can track meta-information about reasoning and decision making as they work. A theorist might, for instance, tag starting judgements as #initialcommitments, changing those tags to #discarded, #replaced, or #adjusted as and when they are supplanted (cf. Baumberger & Brun, 2021, p. 7932). Or, they might tag the #rationale for each of these decisions in a separate note, linking between these justifications and the elements they explain. Thus, in the chain of notes a₁ - b₁ - c₁, note b₁ could explain why a₁ was retired in favour of c₁. Analysing these justificatory notes in aggregate will, in turn, likely help theorists to summarise their overall epistemic goals and favoured theoretical virtues. Any notes recording such higher-level justifications will likely garner significant numbers of links and so become focal points on the graph, further attesting to their importance in the equilibrating process.

Ascertaining similar levels of insight, let alone gathering quantitative evidence akin to thickets of links, strikes me as hard to achieve with any certainty without working digitally. In this way, the explainability of an RE is enhanced by its digitisation. Digitising effectively makes all of these connections and reasons explicit, forcing reasoning out of the theorist’s head and into their (digital) notes. Just as writing helps to clarify our ideas, constructing DRE will thus help to clarify the reasons for the interconnections between those ideas. This has obvious value for theorists, as it puts them in a better position to be able to explain to themselves why they made certain decisions with reference to their web of notes. However, when it comes to evaluating reasons and justifications, digitisation potentially has far more to offer to people other than the theorist. The real value of DRE comes from their potential to trigger significant reforms around research dissemination and consumption, providing a far more straightforward way for readers and critics to explore a theorist’s entire position.

Fortunately, a solution is already close at hand for the theorist working digitally. Most obviously, digital information is endlessly editable, meaning there is no need for the fixity of traditional publishing. Embracing DRE would therefore empower theorists to work flexibly—allowing them, for instance, to add new branches to stable and coherent positions at any time or, more ambitiously, even extensively alter existing branches and clusters of reasoning in light of new evidence. For transparency and explainability’s sake this would all need to be tracked and recorded, but anything from individual principles and judgements all the way up to entire equilibria could, in theory, be readily edited or even outright replaced at will. Rather than calcifying their equilibria for print, digitally-minded theorists could thus continually revise and share works in progress, or even iterate upon published DRE after the fact, better reflecting Rawls’ notion of an indefinite struggle for RE.

More importantly, digital information is also endlessly reproducible; data can be non-destructively copied and shared for next to no cost. This utterly nullifies the need for concision that printing presses once necessitated. Some of this benefit has of course already been felt in academic publishing with the advent of online access journals and publications, but it would also render the cheap and easy distribution of entire equilibria eminently possible. Indeed, I think theorists could and should be publishing DRE online—preferably via agreed upon, open formats, held in open data repositories—and explicitly referencing them in any publications. Under this new norm, rather than reconstructing a (presumably) coherent equilibrium from mere excerpts, readers would instead be able to explore the total coherence of a position, should they so wish. This would allow them to bypass all ambiguities and refer directly to the relevant RE to contextualise a theorist’s arguments and understand their wider commitments⁸. Such radical transparency would drastically increase the clarity of RE-based discourses, bringing the benefits of externalisation, retraceability, and systematicity—not to mention graph visualisation—to the wider academic community.

Indeed, successful prior experiments with digital zettelkästen in other contexts provide reasons to think this could be eminently feasible, albeit with the caveat that they have so far all been constructed for non-RE purposes. Numerous successful projects to develop so-called ‘second brains’, ‘digital gardens’, and ‘digital vaults’ are currently being maintained using zettelkasten applications and uploaded to the internet for anyone to browse and critique. And many of these comprise thousands of interconnected, tagged, and referenced academic notes. These examples—which I would encourage readers to seek out if they are interested—provide compelling reasons for thinking the same should be possible in an RE context.

Why, though, would a theorist want to share their RE? And why should they care about exploring other peoples’? Quite simply, open distribution is essential for realising a vital yet hitherto neglected aspect of the RE process. To reach a wide RE, theorists should be evaluating their own position against others’—especially those with distinct experiences and outlooks—to alert them to their biases as well as any under-theorised areas of thought (Daniels, 1996, p. 2). Yet, as Baderin (2017, p. 14) notes, whilst this notion of RE comparison is “commonly put forward” in the literature, it has so far “not been adequately developed” in practice. This is surely due in part to the sheer practical impossibility of making such comparisons under the status quo. Not only are most equilibria currently never published but, even if they were, print-based distribution would be inadequate for all the reasons listed above. By contrast, a standardised system for publishing DRE would enable direct comparison, addressing one of the biggest current barriers to implementing wide RE⁹.

Of course, I can only suggest potential models for distributing DRE here. The wider community will need to agree on an ultimate approach. But perhaps the least inspired solution would be for the established journal publishers to host DRE in online appendices accompanying traditional publications. Sections of essays and monographs could then link through to the relevant portions of a theorist’s digital RE, hugely increasing explainability. This is certainly an option. But if the whole point of wide RE is that theorists should be exposed to as many other viewpoints as possible, then to me that should be reason enough to reject locking DRE away behind the paywalls of academic publishers. Instead, the alignment of DRE with an open science ethos, which holds that research processes and outputs should be kept “transparent and accessible” and be “shared and developed through collaborative networks” (Vicente-Saez & Martinez-Fuentes, 2018, p. 434), seems a more natural fit.

What would an open approach look like? Given that DRE are essentially folders of plain-text notes, they will be relatively easy to share. Whilst theorists could therefore freely host DRE on their personal sites, I suspect many would rather focus on their theoretical work. To nonetheless facilitate an ‘open humanities’ or ‘open scholarship’ approach (Knöchelmann, 2019), a better option would be to utilise a free or not-for-profit service to manage hosting and any technical issues on theorists’ behalf. Inspiration could be drawn from numerous preprint archives, or else specialist services like PhilPapers (which runs free research indexing and archive services for philosophers) or the more general GitHub (a popular open-source software host) could provide avenues for transparent and accessible DRE dissemination. Alternatively, universities and other research institutions could extend the scope of their archives and open data repositories to store DRE, too.

Whatever form distribution might take, a commitment to open access DRE can be expected to bring the benefits of the former to the latter. This would not only remove existing barriers to accessing RE-based knowledge, increasing research impact by exposing more potential interlocutors to theorists’ research, but also possibly expand the range of voices that can take part in academic discussions (Willinsky, 2006, p. 32). Rather than being locked away or limited to the ivory tower, open access DRE would be widely and freely distributed over the internet. This would maximise opportunities for RE comparison, bringing with it all the chances for mutual learning and the strengthening of arguments that that would entail. This is in direct support of the aims of wide RE and, further, potentially even hints at the beginnings of a method for identifying those principles of justice that could support an overlapping consensus of the kind that emerged in the later Rawls.

This also marks a turning point in my argument. Until now, I have largely adopted the point of view of the individual theorist. Any benefits for the broader research community have been incidental—mere by-products of reconstruction or the greater flexibility and explainability of externalised equilibria. Implicitly, the information flow has been one-way, from theorists to their readers, much like under the print-based dissemination model. But fully embracing open scholarship would make it entirely possible to conceive of two-way flows between theorists and their readers. It is hard not to understate the possibilities this could unlock. A digital-first approach to theorising could radically alter how theoretical work is carried out by involving others in the RE process. It is to these possible social aspects of theorising that I will now turn.

If we accept that no theorist is an island, practically judging coherence (and other properties) becomes far more viable. An analogy can be made to existing processes of academic peer-review. With regards to the challenges from Strong and Van de Poel, imagine if theorists could upload their DRE to open access repositories to solicit communal evaluation from colleagues and peers. Third parties could then explore these equilibria and offer their opinions about whether more work is required to reach RE by submitting annotations and alterations. Such critical scrutiny would help theorists ascertain whether they had included all necessary elements in their RE, whether they had overlooked any inconsistencies, as well as whether each individual element was in receipt of enough support. In the spirit of wide RE, theorists would then be well-placed to go away, deliberate, and decide whether or not to incorporate any critical suggestions back into their overall position.

Such communal evaluation should be distinguished from other attempts to bring third parties into the RE process. Pioneering theorists including Miller (2003), Martine de Vries and Evert van Leeuwen (2009), as well as Jonathan Wolff and Avner de-Shalit (2007), have all sought to include others in their theorising to various extents. But such efforts have mostly involved treating public opinion as an early input to RE; a source of initial judgments and/or principles. Whilst some have gone further, proposing to involve multiple people in RE’s balancing and revisioning processes (cf. Savulescu et al., 2021, pp. 657–659), I agree with Baderin (2017, p. 12) that this seems unworkable in practice. Garnering public input to an RE may well be desirable and achievable, but notions of large-scale collective theorising seem to underappreciate just how difficult—if not impossible—it would be to actually reconcile nuanced, often-conflicting worldviews into a mutually acceptable and shared whole.

To clarify, I am not suggesting that theorists should collaborate with their peers to reach collective DRE. Peer-reviewers would not strive for a shared consensus, and nor would they need to internalise any equilibria they evaluated. They would not even need to agree with the theorist’s choice of inputs, though should probably propose any overlooked ideas for consideration. Instead, my proposal here is best understood as an attempt to split the task of RE evaluation up into more manageable chunks. This is intended to help deal with the complexity and scale of DRE. At the micro level, for instance, justice theorists could draw on their expertise to scrutinise any justice-related elements of a position, whilst metaethicists could focus on areas relevant to their own proficiencies. At the macro level, we could then aggregate all these judgements to help ascertain which equilibria were more or less coherent overall. Though care would evidently need to be taken to avoid reductionism, I nonetheless contend that this communal approach offers a potentially workable method for evaluating coherence. Determining the consistency of a local cluster of related ideas (and their various interconnections) seems far more viable for individuals than judging global coherence.

Naturally, however, any chorus of critical voices is unlikely to be entirely harmonious. Peer-review will thus provide the theorist with many considerations to reflect upon, some of which may even conflict. But this is not an issue. As they retain ultimate control over their position, the theorist will simply need to decide what to accept and integrate into their RE, or else provide reasons (perhaps publicly) for rejecting critics’ suggestions. Much like with the manuscript peer-review process, such efforts would amount to a form of loosely democratic quality-control (Knöchelmann, 2019, p. 9). Compared to the status quo, where there is limited transparency around reflective equilibria and almost no scope for their critique, the potential gains should be clear. Even if all errors, inconsistencies, and insufficiently supported elements are not uncovered, exposing DRE to far more critical scrutiny can be expected to increase their quality.

That said, there is no need to completely foreclose all collaborative possibilities. Whilst I personally doubt how effective collective theorising could be, these proposals do provide scope for adventurous theorists to explore aspects of cooperative working—a by-product of the greater accessibility and flexibility of DRE. Under the open access model, two or more theorists could plausibly come together and work towards a shared RE, much like multiple authors collaborating on a Google Doc. As mentioned earlier, though, the likelihood of two theorists completely agreeing on all the necessary judgements and principles, not to mention mutually accepting all revisions, strikes me as vanishingly improbable. But such prospects could nevertheless pose an interesting avenue for future research to explore.

Perhaps a more realistic possibility is that third parties could branch off from a theorist’s RE to create derivative positions where there is divergence in only some aspects of thought. Each theorist would then still be responsible for managing their own RE. A student might largely agree with their professor, for instance, but wish to change one important principle. After duplicating their professor’s position, they would then need to refactor any ideas which the new principle touched upon. This would likely require significant knock-on alterations to render the RE coherent again. Along similar lines, when encountering compelling alternative views we might imagine merging portions of a theorist’s position into our own¹¹. Lacking expertise in jurisprudence, I could defer to a more qualified theorist on matters relating to the law, incorporating their insights into my own RE. In this way, working digitally could potentially enable us to ‘remix’ new equilibria.

However, feasibility concerns aside, I am reticent to fully endorse these more radical proposals. An undesirable corollary of publishing DRE in this level of detail is that it will also make it easier to plagiarise views. Academics have, over centuries, developed robust norms for acknowledging the ownership of ideas, but we are already seeing these practices strain in the globalised knowledge economy. Any developments, like DRE, that might make it easier for unscrupulous individuals to pass ideas off as their own thus need to cautiously implemented. Even where there is no malicious intent, like with my jurisprudence example, mechanisms for tracking intellectual property would need to be properly engineered. Solutions would thus require the cooperation of zettelkasten software developers, who would need to build ways of tracking authorship metadata into their applications. This may not be an insurmountable issue, but will need to be tackled if DRE are to proliferate.

If software developers are interested in altering their applications to better support DRE, there is one other modification I would also implore them to make. Tracking the development of ideas over time is currently a challenge in existing tools. I have suggested that, as notes are created, edited, and replaced, these changes and their justifications need to be recorded. Without such evidence, it is hard to evaluate how a theorist’s position has evolved and developed. Something akin to the ‘version control’ systems that track code revisions in software development would thus be useful here. Embedding similar systems within programs designed for constructing DRE could help ensure that any changes, and the reasons for making them, can be understood by third parties once an RE is published. It would also help theorists to manage their DRE as they develop, affording them yet more ways to clarify their thinking.