I. Introduction
Emerging technologies are widely expected to have a profound impact on corporate governance. Distributed ledgers and blockchain technology may pave the way towards more direct shareholder involvement in governance by facilitating coordination and reducing costs associated with voting and trade clearance, while artificial intelligence may shrink the decision-making terrain where corporations remain reliant on human management.Footnote 1 Meanwhile, commentators observe organisational trends towards “decentralization”, as firms adopt “more direct and ‘flatter’ forms of interaction and transaction”, and “disintermediation”, as we become “less trusting of organizations, authorities or procedures [and] instead place our trust in machines or algorithms”.Footnote 2 Over time, it is reasonable to expect that new corporate governance practices may emerge – and that various existing corporate governance structures may become more or less significant – in response to such developments.Footnote 3
Nevertheless, discussion of these emerging technologies and their anticipated corporate governance impacts may have gotten well ahead of itself. As a descriptive/predictive matter, Luca Enriques and Dirk Zetzsche point to what they call the “tech nirvana fallacy” – the tendency to posit “perfect machines” as an alternative to flawed human beings and market actors, in which light a steady march towards the former and away from the latter appears inevitable.Footnote 4 In reality, there is ample reason to be sceptical about the capacity of emerging technologies to facilitate a high degree of coordination among shareholders or to substantially displace human managerial judgment.Footnote 5 Meanwhile, as a normative/prescriptive matter, the “technolibertarianism” that characterises coding cultureFootnote 6 dovetails with the economic libertarianism that characterises prominent conceptions of corporate governance,Footnote 7 such that following the technology wherever it leads and accepting the corporate governance consequences that follow appear straightforwardly desirable.Footnote 8 In fact, we need not do so; even if, in theory, emerging technologies were to develop to the point that their most enthusiastic proponents predict, it would remain entirely within our purview to accept all, some, or none of their novel capabilities based on other policy grounds and regulatory imperatives.Footnote 9
This article examines the impact of emerging technologies from both positive and normative perspectives, focusing on how such developments may impact debates about corporate purpose in the context of publicly-traded companies – which typically boil down to how much governance power shareholders should have and the degree to which boards of directors should be required to focus on shareholders’ interests.Footnote 10 While the relevant technology may be new, these fundamental normative issues are not – and I argue that it remains incumbent on us to evaluate potential governance innovations critically, through conceptual lenses that are already well familiar to the field.
I direct the analysis towards corporate governance in publicly-traded companies because that is the context where debates about corporate purpose are thought to be most consequential; such issues tend not to loom so large in the closely held context, where equity holders often participate directly in management.Footnote 11 It bears emphasising that this framing also has the effect of bracketing some of the more extreme and speculative technological scenarios discussed in the literature. For example, publicly-traded corporate structures governed solely and directly by shareholders would appear technically challenging and essentially implausible without some distinct response to rational apathy, which has yet to be identified.Footnote 12 Likewise, there is reason to doubt that companies governed solely and directly by algorithms would prove workable, or attractive to the general investing public, anytime soon.Footnote 13 A corporate structure involving the familiar power constituencies – shareholders and boards of directors – will likely persist, and the article accordingly focuses on how emerging technologies might impact debates about the balance of power between these constituencies, and the appropriate metrics of board decision-making.Footnote 14
While limited shifts along the foregoing lines – in favour of greater shareholder involvement, with diminished reliance on human management – are conceivable and might prove desirable, I argue that any such changes to corporate law and governance ought to be closely scrutinised and defended first, in familiar normative terms. The article begins, in Section II, with an overview of emerging technologies that could augment shareholder governance capacities – notably distributed ledgers and blockchain technology – and/or diminish the need for board monitoring and judgment – notably artificial intelligence. Section III then provides an overview of corporate purpose debates and assesses these technologies through those familiar lenses, concluding that their potential impacts are complex and varied – meaning that we have choices to make, which will have to be grounded in some normative conception of corporate purpose external to the technology itself. Section IV concludes.
II. Shareholders, Boards and Technology
New technologies are widely expected to alter the governance of publicly-traded companies. This part of the article provides an overview of two significant sets of developments: first, those that could augment the capacity of shareholders to organise and exert themselves in corporate governance, notably distributed ledgers and blockchain technology; and second, those that could diminish the need for human beings to perform various monitoring – and perhaps even strategic – roles traditionally performed by boards of directors, notably artificial intelligence.
A. Distributed Ledgers, Blockchain and the Role of Shareholders
The current architecture for trading and voting securities is widely understood to be a mess. Fungible bulk shareholding emerged as a response to the inefficiency of transferring physical stock certificates, but it relies on a cumbersome array of intermediaries, including banks, brokerage houses, and various service providers, to transmit trading and voting instructions back and forth between the depository holding the stock and the beneficial investors – a process requiring a complex array of proxies and contractual arrangements.Footnote 15
This process imposes substantial costs and remains rife with potential for error, creating significant legal problems if investors’ instructions are not properly carried outFootnote 16 – complications vividly illustrated by disputes arising in the peculiarly active US corporate litigation environment.Footnote 17 In his In re Appraisal of Dell Inc. decision, Vice Chancellor Travis Laster of the Delaware Court of Chancery (the premier US business courtFootnote 18) held that T. Rowe Price was not entitled to an appraisal remedy following a merger that they opposed – because their voting instructions were lost in the shuffle and, contrary to their wishes, their shares were voted in favour of the merger.Footnote 19 The error cost T. Rowe Price $200 million.Footnote 20 More generally, it is thought that outcomes of close elections in corporations with fungible bulk shareholding are essentially unverifiable – a reality that, according to Laster, “creates doubt about the integrity of the stockholder vote” and may “undermine the legitimacy of our corporate governance system”. According to one prominent Delaware practitioner, where voting margins are narrower than 55 to 45 per cent., “there is no verifiable answer to the question ‘who won?’”Footnote 21
Distributed ledgers and blockchain technology might alter this landscape for the better. In contrast with traditional ledgers, a distributed ledger involves “distributed computer software running on peer-to-peer networks that offers a transparent record management system in which transactions are cryptographically signed and validated with timestamps using a consensus mechanism” – a system that inspires trust precisely because “no single entity, person, or computer node owns, controls or otherwise maintains [it]”.Footnote 22 Blockchain, then, refers to the protocol through which “data bundles” are stored “in a strict time-related series which links each block, through a time stamp, to the previous and subsequent blocks”, creating a permanent chronological record of transactions.Footnote 23 In turn, distributed ledgers and blockchain technology provide the foundation for so-called “smart contracts” – software permitting an underlying legal contract to be made self-enforcing to the degree that the parties’ agreement can be reflected in the code itself as a series of if-then instructions. Critically, such a system may eliminate the need for trusted professional intermediaries to the degree that the parties trust the code itself (and, implicitly, the coders).Footnote 24 These technologies were first used to support digital currencies,Footnote 25 and “[o]rganizations in financial services, health care, real estate, as well as governments are exploring or have already implemented blockchain solutions”.Footnote 26
Proponents of these technologies anticipate that they will bring greater accuracy and transparency to voting and clearance, permitting more direct shareholder involvement in corporate governance at lower cost by dispensing with the complex chain of intermediaries described above,Footnote 27 although this is expected to require some degree of centralisation. An “unpermissioned” ledger for a digital currency (e.g. Bitcoin) is built “to allow anyone to contribute data to the ledger and for everyone in possession of the ledger to have identical copies”, maintaining the ledger's integrity through a complex and costly consensus process that precludes any single actor from controlling additions to the ledger.Footnote 28 A “permissioned” ledger, on the other hand, may be owned by a single entity or group that controls admission to the network, and the ledger's integrity is maintained through “a limited consensus process” that is “carried out by trusted actors”.Footnote 29 In such a system, “the parties are not looking to the ledger technology itself to generate trust”, but rather use it “for its other characteristics, such as the speed and efficiency of on-chain transactions” carried out through centralised consensus mechanisms that are generally “faster, cheaper and more scalable”.Footnote 30 The participants in a permissioned ledger are known to one another and are likely bound by a distinct agreement governing participation, which “facilitate[s] easier interaction with traditional dispute resolution procedures and enable[s] compliance with regulatory requirements”.Footnote 31 While such an arrangement sacrifices open-source scrutiny, “thereby reducing opportunities for mistakes to be identified and remedied”, operational efficiencies and regulatory imperatives have prompted financial institutions and governments alike to focus their efforts in this direction.Footnote 32
Although distributed ledgers and blockchain technology remain “at a very early stage of development”, and “the technology of smart contracts is still in its infancy”,Footnote 33 these developments have been endorsed by public actors around the world, and substantial initiatives to develop and apply such technology are underway. In his In re Dole Food Co. opinion, Vice Chancellor Laster observed that “facially valid claims” to settlement proceeds following a contested merger were submitted for a number of shares far exceeding the actual shares in the class, yet explained that “[w]ithout obtaining detailed records about the millions of trades that took place during the three days leading up to the closing, it is impossible to determine who owned the shares as of closing” – rendering it “functionally impossible to resolve the share discrepancy in a practical or cost-effective manner”.Footnote 34 Laster resorted to distributing the settlement proceeds to the depository, placing the onus on them to figure it out.Footnote 35 At the same time, he observed that distributed ledger technology “offers a potential technological solution by maintaining multiple, current copies of a single and comprehensive stock ownership ledger”, and cited the State of Delaware's support for the development of such technology.Footnote 36 Laster refers to the Delaware Blockchain Initiative, announced in 2016,Footnote 37 which resulted in amendments to the Delaware General Corporation Law permitting the use of distributed ledger and blockchain technology for corporate records, including the stock ledger.Footnote 38
Meanwhile, substantial efforts are taking shape in the EU and the UK. The European Commission has identified complex intermediary structures as a hindrance to the effective exercise of shareholders’ rightsFootnote 39 and, in 2018, launched the EU Blockchain Observatory and Forum “to monitor trends and developments”.Footnote 40 The UK government, meanwhile, has actively explored opportunities to develop and apply such technologies, including through Digital Catapult, a UK government agency that provides facilities, networking, and resources to promote business adoption of various new technologies, including distributed ledgers and blockchain technology, as well as artificial intelligence.Footnote 41 Such efforts have been motivated in part to preserve London's leadership position in global finance,Footnote 42 and potential applications in financial markets have attracted attention on both sides of the Atlantic.Footnote 43
The disruptive potential of such initiatives is perhaps most clearly reflected in the fact that the traditional intermediaries themselves are scrambling to establish a foothold in this new technological terrain – an effort presumably aimed at maintaining relevance in a future that otherwise threatens redundancy.Footnote 44 Securities depositories may not be alone, however, as some even foresee the eclipse of the board of directors itself. Taken to their logical extreme, technological advances enhancing the capacity of shareholders to participate in corporate governance might facilitate greater reliance on partnership-style governance structures in which equity holders directly manage the business. While not unknown to corporate law as a doctrinal matter,Footnote 45 the idea is that this might be done at a much larger scale via the blockchain. A prominent effort towards this end was the DAO (i.e. Decentralised Autonomous Organisation) created by Slock.it in 2016.Footnote 46 While not formally incorporated, the organisers clearly envisioned “an organization similar to a company”,Footnote 47 albeit with some crucial differences. The DAO was designed to operate as “a leaderless, decentralized venture capital firm that would allow investors to vote on and collectively fund proposals” through smart contracts running on the Ethereum blockchain platform, effectively “substituting code for the directors and officers”.Footnote 48 The undertaking was phenomenally successful (at least initially), raising over $168 million.Footnote 49
In assessing the potential corporate governance impacts of such technology, it is important to recognise that substantial legal, technological, and market impediments remain. For example, with distributed ledger and blockchain technology, the very fact that the record cannot be amended is problematic in light of the “right to be forgotten” under the EU's General Data Protection Regulation.Footnote 50 Smart contracts, meanwhile, effectively require that the parties’ agreement be reducible to if-then instructions, which is not strictly feasible in any but the simplest circumstances.Footnote 51 Even with more sophisticated software able to accommodate greater complexity, however, enforcement problems arise with “black box” algorithmic contracts increasingly relying upon “decision-making procedures that are not functionally human-intelligible before the program runs – and often cannot even be parsed after the program runs”.Footnote 52
At the same time, there is reason to doubt that widespread adoption of decentralised blockchain technology will prove beneficial or cost-justified relative to existing or nascent alternatives. The benefits of a truly decentralised system “come at a great cost” – notably, participants’ incentives have to be carefully calibrated, and “malicious users” cannot straightforwardly be excluded; maintenance costs “are orders of magnitude higher” due to the redundancies built into a distributed method of establishing consensus; a “forced upgrade” generally cannot be made because participants have no obligation to accept it; and scaling is likewise difficult because the data “has to live in hundreds or thousands of places”.Footnote 53 The cumbersome distributed consensus mechanism likewise tends to render such systems quite slow, prompting comparison to the “Volvo diesel: it is sturdy and safe, but certainly not very comfortable and fast”.Footnote 54 These costs have prompted further experimentation with alternative distributed ledger designs that employ simpler consensus mechanisms, essentially trading off safety and immutability for greater speed and scalability,Footnote 55 as well as consideration of centralised alternatives including permissioned blockchain systemsFootnote 56 and various means of achieving auditability and backup benefits through traditional databases (e.g. receipts and cloud-based solutions).Footnote 57 Decentralised blockchain technology certainly cannot be declared “obsolete” given ongoing development efforts and the advantages of safety and immutability,Footnote 58 but the extraordinary speed and volume of trading in securities markets suggest that centralised approaches of one form or another will likely prove most attractive in that context.Footnote 59
Meanwhile, experience to date suggests that the need for human leadership remains acute when something goes wrong in organisational settings reliant on such technology – as the collapse of the Slock.it DAO illustrates. In June 2016, “an unknown person withdrew ether worth approximately $55 million from The DAO by exploiting flaws in the code” – an action apparently permitted under the terms of the smart contract governing the undertaking, yet obviously not contemplated by the organisers and other participants.Footnote 60 An anonymous open letter, purportedly from this individual and addressed to “the DAO and the Ethereum community”, defended the action as “fully compliant with United States criminal and tort law” and a straightforward consequence of an “explicitly coded feature” of the relevant smart contract. Accordingly, the letter argued that any effort to reclaim the ether would undermine confidence in such undertakings.Footnote 61 Ultimately, a so-called “hard fork” effectively undid the transfer with the support of a large majority of participants,Footnote 62 and in so doing, it “accomplished two things that were supposed to be impossible on a public blockchain: rewriting the history of transactions, and introducing human intervention to negate the unanticipated consequences of a self-executing smart contract”.Footnote 63 The hard fork “created what is now Ethereum; Ethereum Classic is the original unforked version of Ethereum”.Footnote 64
The DAO episode suggests that leaders capable of reacting quickly to unexpected circumstances remain desirable in undertakings of any scale.Footnote 65 In this light, recall that financial institutions and governments have focused on developing more centralised systems precisely to preserve control, improve regulatory capacity, and reduce operating costs.Footnote 66 At the same time, this episode correlatively points towards an enduring limitation on the governance capacities of minority equity holders in any form of organisation – rational apathy. Technology that reduces transaction costs associated with voting might mitigate this problem,Footnote 67 but other causes of this pervasive phenomenon would remain unaffected, including free riding and costs associated with acquiring the expertise required to vote in an informed manner.Footnote 68
The foregoing challenges are considerable and suggest that fully decentralised trade clearance or business organisational structures are unlikely to take shape in the near-term. It certainly remains conceivable, however, that forms of distributed ledger and blockchain technology could facilitate greater involvement by shareholders in corporate decision-making to the extent that costs of coordination and voting could be meaningfully reduced. Likewise, pro-shareholder corporate governance reforms might find greater political support to the extent that such developments undercut opposition based on costs and delays historically associated with cumbersome intermediated mechanics for voting and clearance.Footnote 69 As discussed below, such a shift in favour of shareholders could occur even if heavy reliance on distinct management remains the norm, further invigorating debates regarding the optimal balance of power and degree of focus on shareholder interests in corporate governance.Footnote 70
B. Artificial Intelligence and the Role of Management
Boards of directors have traditionally performed two core functions – monitoring corporate performance and formulating corporate strategy.Footnote 71 To the degree that decision-making relevant to these functions could be automated, however, some anticipate that artificial intelligence might reduce the need for human management, in turn reducing costs and improving accuracy and efficiency.Footnote 72
At its core, artificial intelligence represents a set of statistical and algorithmic tools, so the moniker “intelligent” may, strictly speaking, be a misnomer.Footnote 73 Generally speaking, however, the term refers to “a set of techniques aimed at approximating some aspect of human or animal cognition using machines”. Attracting particular attention are those techniques referred to as machine learning, which involves “the capacity of a system to improve its performance at a task over time”, and deep learning, which involves the use of “many-layered structures to extract features from enormous data sets in service of practical tasks requiring pattern recognition”.Footnote 74 Such technology is widely expected “to affect all operational as well as internal control matters, from strategy setting to risk management and compliance” – both because of its capacity to draw upon a far broader range of data than human decision-makers can, and because its proponents believe (unrealistically, as I discuss below) that artificial intelligence will transcend human biases and social dynamics that impair both individual and group decision-making.Footnote 75 In terms of how it might interact with human decision-makers, artificial intelligence may be categorised along a spectrum from “assisted” artificial intelligence, which supports human decision-makers by performing limited tasks (e.g. Apple's Siri); to “advisory” artificial intelligence, which likewise “does not replace human intelligence but leverages or improves it, such as by giving information and advice that would otherwise be unavailable or more difficult and time consuming to obtain” (e.g. IBM's Watson); to “autonomous” artificial intelligence, which literally involves the machine itself becoming the decision-maker (e.g. a self-driving car).Footnote 76
As this spectrum towards increasing technological dependence implies, many imagine a future in which machines displace human decision-makers entirely – including in the realm of corporate governance, where some see real potential for exclusive reliance on algorithmic decision-making. According to an Accenture study on the effects of artificial intelligence on corporate management, “[s]ome of our interviewees questioned whether the manager role as we know it will survive”Footnote 77 – an outcome that might be described as the logical, if ironic, extreme of Taylorist scientific management.Footnote 78 Indeed, while corporate laws typically assume that directors will be human beings, and in some cases expressly require this,Footnote 79 it nevertheless appears to be the case that a completely autonomous business entity – existing solely to provide “housing” for an algorithmic decision-maker, with no ongoing human involvement – is already a legal possibility.Footnote 80
As with the technologies discussed above, substantial legal, technical, and market concerns render it difficult to predict how this might play out, but counsel scepticism regarding the more extreme claims along these lines. Once again, “black box” algorithmic outcomes not susceptible to straightforward explanation create numerous legal and regulatory complexities in wide-ranging contexts,Footnote 81 and there remains significant legal uncertainty regarding who ought to be liable when algorithms run amok.Footnote 82 Other concerns include consumer data protection,Footnote 83 inability to model truly novel events and circumstances,Footnote 84 industry dominance (relative to governments and academia),Footnote 85 and replication of discriminatory biases.Footnote 86 Despite the patina of objectivity cultivated by purveyors of this technology,Footnote 87 a growing literature documents the substantial degree to which biases embedded within the underlying data, and/or within assumptions that coders bring to their work, can skew the outcomes of algorithmic decision-making in ways that are unjust, inefficient, or both.Footnote 88 While such dynamics are as old as actuarial science, these problems become more difficult to identify and address “as large data sets and more-complex models become widespread”.Footnote 89
Meanwhile, some who anticipate completely autonomous legal entities fear that, if they go unchecked, the result will be a dystopian hellscape in which marauding algorithmic corporations perpetrate all manner of crimes and mayhem, presenting a fundamental threat to our very survival. Lynn LoPucki, for example, argues that algorithmic entities’ (AEs’) “greatest comparative advantage would be in criminal enterprise. Because they lack human bodies, AEs are harder to catch and impossible to punish. AEs need not fear death or capture. They can replicate themselves without ego and sacrifice themselves without motive. They need not recoil at the necessity to do violence to humans”.Footnote 90 Even more alarmingly, Stephen Hawking fears that “AI may replace humans altogether”.Footnote 91
These are weighty concerns indeed, and might well prompt substantial direct regulation of such technology,Footnote 92 but the current state of development and the inherent limitations of current approaches generally suggest that both our worst fears and our most optimistic expectations are overblown. There is a “wide range of estimates of the time horizon until the advent of [artificial general intelligence] – from a decade to two centuries”,Footnote 93 but the prevailing view among “the mainstream AI community” is that “we are hundreds of years away from understanding how to create machines capable of formulating an intent to harm”.Footnote 94 Meanwhile, the inherent limits of what current approaches can deliver are increasingly recognised and acknowledged. While deep learning, “the best known technique”, has “yielded numerous state of the art results, in domains such as speech recognition, image recognition, and language translation”, Gary Marcus explains that it cannot be expected to provide “a general solution to artificial intelligence”.Footnote 95 As “a statistical technique for classifying patterns, based on sample data”,Footnote 96 there are several inherent constraints on the capacity for such systems to replicate the subtleties of human judgment. Notably, their strict reliance on labelled training examples renders such systems incapable of formulating or learning from “abstract relationships”, applying general concepts to “scenarios that differ in minor ways from the ones on which the system was trained”, grappling with hierarchical structures (such as natural language syntax), adjusting to dynamic circumstances, or drawing novel “open-ended inferences based on real-world knowledge” – all of which in turn render such systems “easily fooled”.Footnote 97 Overall, deep learning “works great on problems like speech recognition in which there are lots of labelled examples, but scarcely any even knows how to apply it to more open-ended problems”. Marcus elaborates that tasks having “less to do with categorization and more to do with commonsense reasoning essentially lie outside the scope of what deep learning is appropriate for”, and that this technology accordingly “has little to offer such problems”.Footnote 98 As corporate managers are routinely called upon to make difficult judgments in dynamic contexts involving a high degree of uncertainty,Footnote 99 the foregoing limitations render it implausible that artificial intelligence could outright replace human decision-making in corporate governance anytime soon. As Christopher Markou and Simon Deakin observe, human beings remain “far more efficient at learning complex rules and generalising abstraction relationships than [deep learning] systems”.Footnote 100 Accordingly, public investment markets will likely continue to favour reputable human leaders as a basis for trust in corporate governance,Footnote 101 with the replacement of human directors by algorithms more likely to arise in sharply constrained settings – for example, “in the context of subsidiaries with very narrowly-defined domains of operation”.Footnote 102
It remains conceivable that such technology could, in more limited ways, reduce reliance upon human decision-making for certain board-level tasks – most obviously with monitoring, and potentially with advising on the development of strategy as well. It is widely thought that administrative tasks will gradually shift to artificial intelligence, yet increasingly acknowledged that “soft” skills requiring social and emotional sensitivities, and judgment in the face of radical uncertainty, will remain the province of human decision-makers.Footnote 103 We are already witnessing growing reliance on artificial intelligence-based tools at the operational level, particularly in financial services – so-called “FinTech”,Footnote 104 including “robo-advisers” and the likeFootnote 105 – and it is not difficult to envision directors drawing upon similar tools in an organisation's own governance.Footnote 106 In this light, giving an algorithm a nominal board seat with non-voting observer status, as some entities have done, represents a splashy gesture in this direction – demonstrating how artificial intelligence might eventually create heuristics and defaults that frame, and therefore condition, human decision-making in corporate governance,Footnote 107 while underscoring how far we in fact remain from actual robo-directors exercising independent managerial discretion.
III. Technology and Corporate Purpose
In discussions of rapidly developing technologies like those described above, the line between what could be and what ought to be seems peculiarly predisposed to blur. Technological advances – including potential applications in corporate governance – take on a patina of inevitability, with law and regulation implicitly placed in a reactive posture. As noted above, this likely reflects the confluence of technological libertarianism and economic libertarianism that merge and reinforce one another in this distinctive context. Given the stakes involved, it is all the more important to recognise this tendency in the discourse on technology and corporate governance and to examine associated normative issues in a deliberate way.
This part of the article undertakes such analysis in the context of publicly-traded companies, focusing on debates about how much power shareholders should have, and how much discretion boards of directors should have to accommodate interests and values that potentially diverge from what shareholders would prefer in the near-term. In the context of publicly-traded companies, as discussed above, we can safely set aside the more extreme technological scenarios that do away entirely with human boards of directors. Accordingly, following a brief overview of corporate purpose debates, this part evaluates how the foregoing trends and emerging technologies might impact how we think about corporate purpose in publicly-traded companies – and, conversely, how prevailing conceptions of corporate purpose might impact the desirability of pursuing certain technological possibilities as a matter of policy.
A. Corporate Purpose Debates
Debates regarding the appropriate level of shareholder control and the legitimate objectives of corporate decision-making have unfolded quite differently in various jurisdictions, and distinctive balances have been struck based on a number of legal, political, economic, and cultural factors.Footnote 108 Indeed, even closely related common-law systems with substantial legal and market similarities – including the UK, the US, Australia, and Canada – have exhibited considerable variation in terms of the degree of decision-making power that shareholders possess and the degree to which boards of directors are compelled to focus narrowly on their interests.Footnote 109 In assessing how technologies such as distributed ledgers, blockchain, and artificial intelligence might impact these balances, it is helpful to bear in mind the fundamental points of normative disagreement. To that end, I briefly recite the core claims advanced by a handful of prevailing theoretical perspectives on corporate law.
Prevailing theories have advanced very different claims about the nature and purpose of the corporation, with significant ramifications for how we think about shareholder power and the legitimate aims of board decision-making. The nexus of contracts theory, as the name implies, presents corporate law as a set of majoritarian default rules designed to reduce the cost of contracting among various corporate constituencies. Proponents of this view claim that governance rules favouring shareholders through reduction of agency costs are essential to attracting capital providers,Footnote 110 and generally assume that other constituencies and interests can be protected through their contracts with the corporation and/or through extra-corporate regulation.Footnote 111 The team production theory, a variant of nexus theory, presents a radically different picture of the corporation in which the board of directors sits at the centre of a “mediating hierarchy” that aims to promote firm-specific investment by a range of stakeholders, all of whom are treated as residual claimants. The board of directors, in this conception, has an obligation not solely to protect shareholders, but “to protect the enterprise-specific investments of all the members of the corporate ‘team,’ including shareholders, managers, rank and file employees, and possibly other groups, such as creditors”.Footnote 112 Shareholder primacy theory, meanwhile, presents a simpler conception of the corporation, with shareholders at its centre. On this view, shareholders are the owners of the corporation, and they hire management to advance their interests through a straightforward principal-agent relationship.Footnote 113
Each of these approaches fails to provide a fully compelling account of corporate law. For example, nexus of contracts theory cannot explain the law of hostile takeovers, which (at least in the US) markedly departs from the exclusive focus on shareholder interests. Conversely, team production theory cannot explain voting structures and remedies that really do favour shareholders over other constituencies in many circumstances. And shareholder primacy theory fails on multiple fronts, providing no explanation for deviations from the exclusive focus on shareholder interests – including the US approach to hostile takeovers and the UK conception of “enlightened shareholder value” – and being flatly incorrect as a matter of law in depicting directors as the shareholders’ agents.Footnote 114 While managers of publicly-traded companies certainly face substantial pressures to maximise share price over relatively short timeframes, such pressures emanate principally from securities regulation and capital market practices rather than corporate law itself.Footnote 115 Indeed, it is telling that even those who believe, as a matter of corporate law, that long-term shareholder value represents the singular and exclusive corporate objective generally concede that its desirability and (implicitly) its stability may fairly be questioned in light of failures to contain corporate excesses through extra-corporate regulation – the viability of which tends to be assumed by shareholder-centric theories of corporate law and governance.Footnote 116
For present purposes, it is not necessary to dilate on the strengths and weaknesses of the foregoing theories so much as to recognise the substantial and persistent diversity of views regarding what corporate law and the corporate form fundamentally exist to accomplish, how much governance power shareholders ought to have, and how large their interests ought to loom in corporate decision-making.Footnote 117 In this light, an embrace of greater shareholder power and diminished reliance on human managerial discretion would follow logically from the emergence of new technologies only if one already thought that the normative structure of corporate law favoured that approach and that historical deviations from that approach simply reflected second-best structures adopted in the face of technological limitations. Such assumptions – reflecting a shareholder-centric approach to corporate law and governance predicated upon a principal-agent conception of shareholder-manager relations – appear to motivate those concluding that embrace of such technological developments should be virtually automatic.Footnote 118 This vision of corporate law and governance remains contested, however, and nothing about these new technologies, in and of themselves, necessitates such a shift any more than prior technological advances have.
B. The Relevance of Technology to Corporate Purpose
While I argue above that the most enthusiastic claims for these technologies are overblown, I acknowledge that they could nevertheless impact corporate governance in substantial ways,Footnote 119 and none of the foregoing aims to suggest that technological advances cannot reasonably prompt us to re-think fundamental issues of corporate law and governance. We should not, however, pursue whatever the technology would appear to facilitate simply because it is possible, any more than we have done in the past. In the US context, for example, restrictive proxy voting and tender offer rules through federal securities regulation have substantially limited coordination and exertions of power by institutional shareholders to a level well below what even the more limited technology of the twentieth century plainly would have allowed,Footnote 120 and likewise state corporate law has extended only limited voting power to shareholders, relative to other similar common-law systems.Footnote 121 Even in the UK context, however, where shareholders have long possessed direct corporate governance powers exceeding their US counterparts, and where the Companies Act 2006 expressly prioritises shareholder interests, directors typically possess sweeping authority under the articles to determine the course of corporate decision-making and enjoy statutory deference to their own subjective judgment regarding what course of action would most tend to “promote the success of the company” – indeed, they may even “have regard” for other constituencies and values in so doing, an approach termed “enlightened shareholder value”.Footnote 122 In evaluating new technologies with the potential to “affect the balance of power between directors, managers, and shareholders”,Footnote 123 then, the analysis should not take place in a policy vacuum. In tackling these matters, we are called upon to revisit long-standing questions regarding how much power shareholders ought to have and how much discretion boards should have to accommodate competing interests.Footnote 124
The answers in this context are not obvious, despite the air of determinism animating much of the discourse.Footnote 125 The technologies involved are inherently complex, and they interact with one another in complex ways, such that their potential impacts are not always clear.Footnote 126 Additionally, their potential ramifications for corporate governance could be affected by the degree to which the technologies themselves are directly regulated through other fields of law, as some strenuously advocate.Footnote 127 Implicitly, this reflects a recognition that the technology, and even to some degree the underlying science, are themselves social constructions over which we exercise real agency.Footnote 128 Accordingly, one could envision various applications of such technologies that have very different normative implications – some advancing a more shareholder-centric vision of corporate governance, while others might well appeal to self-styled corporate law progressives critical of that approach.Footnote 129
Even if the technology permitted substantial enhancement of shareholder voting powers and capacity to coordinate, it is conceivable that shareholders might remain as passive as they already were due to persistent rational apathy.Footnote 130 To the degree that shareholders remained passive, corporate governance would largely continue, as a practical matter, to default to board discretion. Greater voting power and/or capacity to coordinate – including in pursuit of remedies – might well reinforce activism by a limited category of investors, however, presumably skewing corporate decision-making in the shareholders’ favour.Footnote 131 The likelihood and consequences of such a shift could differ by industry, perhaps raising particular concerns in financial services, where problems associated with shareholder-driven risk-taking in major financial institutions are thought to have exacerbated the financial crisis.Footnote 132
At the same time, however – and even on the assumption of persistent rational apathy – one could imagine such technology being applied in ways that progressives might favour. As a threshold matter, expanded use of blockchain technology in multinational supply chains might provide greater transparency and produce information of interest to various corporate constituencies, improving commercial efficiency while illuminating the degree of compliance with environmental and human rights standards throughout various stages of complex production processes.Footnote 133 To the extent specific physical products could be linked with associated blockchain records – a considerable technical challenge for many types of goods, to be sureFootnote 134 – this could prompt greater attention to such issues among various corporate stakeholders, including beneficial investors in relevant companies, even if an overall tendency towards rational apathy persisted.
Consider, for example, how such technology might affect the voting of shares held by large institutional investors, which directly hold a substantial percentage of equities.Footnote 135 George Geis observes that even if the mechanics of voting were improved by new technologies, such institutions might “continue to cast votes according to a prearranged formula – perhaps hewing to the advice received from third-party proxy advisory firms”, which “could drown out much of the impact of voting with traceable shares in an era when more and more shareholding occurs through intermediaries”.Footnote 136 Proxy advisers are themselves often criticised for simply “mirroring the majority views among institutional investor clients”,Footnote 137 naturally rendering their voting advice highly shareholder-centric. Proxy voting guidelines for pension funds similarly tend to be highly shareholder-centric due to fiduciary duties for plan managers that effectively require not only intense focus on maximising returns to the plan as corporate shareholders, but also engagement in activism.Footnote 138 Geis adds, however, that an “interesting question … is whether it might be possible and worthwhile for individual mutual funds to use distributed ledger technology to ‘pass through’ proportional voting rights to their ultimate investors”. For example, on this approach “an investor who owns 1000 shares of Vanguard's Total Market Index Fund might see this position translate into voting rights for two shares of Apple stock”.Footnote 139
Geis rightly concedes that such an approach could actually increase rational apathy if ultimate beneficial investors were expected to make all such voting decisions themselves.Footnote 140 Yet, intriguing possibilities arise if one imagines potential impacts on the nature of institutional intermediation itself. Specifically, one could imagine using such technology to effectively bifurcate portfolio management intermediation from voting intermediation in various institutional contexts – say, by granting a proxy for one's fractional fund voting power to some distinct firm that undertook to vote the shares in a manner considered broadly consistent with an ultimate beneficial investor's progressive views.
On the management side, meanwhile, it is certainly possible that reliance upon human judgment could be diminished to some degree through increasing voting power and shareholder coordination, on the one hand, and increasing reliance upon algorithmic tools, on the other.Footnote 141 Yet, we would still need human beings to be responsible in some manner for the selection and oversight of the algorithms that animate artificial intelligence-based management tools,Footnote 142 which arguably just shifts familiar problems of corporate governance to a different decision-making context.Footnote 143 While board-level technology committees are not yet engaged with these dynamics, this “would seem to be a natural evolution for the role of tech committees”.Footnote 144 At the same time, as discussed above, it is not realistic to expect that artificial intelligence could effectively displace human decision-makers when it comes to “soft” skills requiring social and emotional sensitivities, and judgment in the face of radical uncertainty.Footnote 145
Moreover, it cannot be assumed that a shift towards algorithmic decision-making, to whatever degree that proves practicable, would necessarily translate to greater shareholder-centrism in corporate decision-making, as that would depend on the substance of the algorithm itself. As discussed above, the algorithm is hardly an ideology-free phenomenon – it invariably reflects values and biases embedded within the coding and the underlying data.Footnote 146 To be sure, one could imagine greater reliance on artificial intelligence militating towards a single maximand to the extent that the technology requires “clearly defined goals and outcomes” to reduce governance-related problem-solving to an algorithm.Footnote 147 It is conceivable, however, that algorithmic decision-making could eventually facilitate a balancing of competing interests to the extent that the technology achieves greater capacity to accommodate complexity. This would straightforwardly undercut arguments in favour of shareholder-centrism resting on the assumption that pursuit of multiple maximands is fundamentally indeterminate and therefore a de facto recipe for managerial rent-seeking.Footnote 148
In any event, increased reliance on algorithmic tools – at all levels of the corporation – raises acute policy considerations. Broader social disruption caused by human redundancies may demand significant attention from policy-makers,Footnote 149 and it is not unreasonable to anticipate that “dramatic changes may generate enough political will to fundamentally change (not necessarily for the better) the corporate structure”.Footnote 150 Meanwhile, growing concern about the degree of industry dominance in the development of artificial intelligence technology that has been achieved by a small number of corporate players – in turn dominated by a small number of people – may itself prompt regulatory attention. Indeed, “as few as seven for profit institutions – Google, Facebook, IBM, Amazon, Microsoft, Apple, and Baidu in China – seemingly hold AI capabilities that vastly outstrip all other institutions”,Footnote 151 and it has not escaped attention that leading artificial intelligence developers “almost all have unusual private sector governance structures designed in general to give control to a small group of insiders with tight connections to each other and very similar backgrounds and interests”.Footnote 152
At the same time, recall that artificial intelligence might itself become the subject of regulation to foster the promotion of broader public values, whether through direct regulation of the algorithmsFootnote 153 or indirect regulation at their point of intersection with relevant markets. The role of securities regulation has received particular attention in relation to distributed ledger and blockchain technology, given that new forms of organisation may involve the sale of securities requiring registration.Footnote 154 At the same time, securities regulation will also presumably require substantial disclosures regarding reliance on artificial intelligence in corporate governance.Footnote 155 More generally, however, there are increasing calls for broad-based regulatory oversight of artificial intelligence systems to ensure that their dogged pursuit of imperfectly crafted objectives does not result in unanticipated bad outcomes.Footnote 156
Another set of dynamics further complicating the landscape as to both shareholders and management – and impacting the degree to which the marketplace actually embraces the technologies discussed above – relates to liability rules and associated exposure. For example, autonomous organisations that do not formally organise as something else will often be considered general partnerships, exposing equity holders to direct personal liability (assuming they can be identified)Footnote 157 – as would autonomous organisations that formally incorporate, yet dispense with boards of directors in favour of direct management by shareholders.Footnote 158
For management, meanwhile, significant questions loom regarding the degree of fiduciary liability that unconflicted directors and officers might face in connection with selection and oversight of algorithms relied upon for corporate governance.Footnote 159 While care and oversight duties have long remained weak, management might come to bear greater responsibility moving forward to the extent that artificial intelligence facilitates the development of more effective corporate monitoring systems at a reasonable cost.Footnote 160 Regardless, however, public company management must still bear in mind the potential for securities fraud suits to impose de facto regulation in such areasFootnote 161 – which would seem to create non-trivial liability exposure, given the substantial disclosures that companies reliant on such technology would presumably have to make for the benefit of investors.Footnote 162 More complex questions yet might arise in the more distant future, should companies managed entirely by algorithms ever become a reality – in which case the acute risk of cost externalisation might spur more drastic responses such as mandatory insurance schemes, direct shareholder liability,Footnote 163 and/or products liability for those producing the algorithms.Footnote 164
In light of the technological developments described above, the innovations they may foster, the potential liability exposure, and the wide-ranging regulatory issues involved, the proliferation of “regulatory sandboxes” in numerous markets around the world is a telling phenomenon.Footnote 165 The aim of a regulatory sandbox is to facilitate the development of both new products and appropriate regulation through a trade-off – innovators can ramp up new products without immediately having to bear the full expense of regulatory compliance (e.g. licencing), in exchange for confining themselves to a limited market under close regulatory supervision for a certain period of time.Footnote 166 To date, this regulatory structure has focused on FinTech innovations,Footnote 167 but one could readily imagine a broader application of the sandbox concept to corporate governance-related innovations more generally. UK regulators have been “proactive in supporting the emerging [distributed ledger technologies] ecosystem” through regulatory sandboxes and other strategies, an approach dovetailing with an overarching “watch and wait” regulatory posture towards this burgeoning field.Footnote 168 In the US, meanwhile, “there is unlikely to be a holistic federal fintech sandbox in the foreseeable future, given the fragmented nature of federal financial regulation”,Footnote 169 but a handful of US states have created sandboxes or pursued more ad hoc efforts to test pilot programs relating to distributed ledger and blockchain technology.Footnote 170 At the same time, regulators are themselves developing and testing similar technologies for distinctly regulatory purposes in response to such market developments – so-called “RegTech”.Footnote 171
Given the likelihood that reliance upon distinctly human decision-making will persist where “soft” skills and judgment in the face of uncertainty are required; the critical role of human decision-makers in selection and oversight of underlying algorithms; and the panoply of regulatory concerns associated with the deployment of such technology, it is unrealistic to expect that artificial intelligence will result in total displacement of human decision-makers in the foreseeable future. Accordingly, it appears most likely that the near-term impacts of artificial intelligence in corporate governance will involve “augmentation” – that is, “putting people first and amplifying what they can achieve”, notably by permitting them to “focus on more value-creating activities, such as those related to judgment work”.Footnote 172 From this perspective, a moderate dose of uncertainty in the specification of objectives might be the order of the day, in so far as it prompts these systems to “ask permission when appropriate”, to “accept correction” and to “allow themselves to be switched off”.Footnote 173 Approached in this manner, the product of managerial decision-making may come to resemble “bricolages of ideas and hypotheses from inside and outside of the enterprise”, drawing on artificial intelligence, but only as one input among many to arrive at what are ultimately human decisions.Footnote 174 The notion of bricolage nicely captures the open-textured nature of a corporate governance system that will likely remain compatible with a number of normative visions that we might choose to pursue.
IV. Conclusion
Emerging technologies likely will, in various ways, prompt us to re-think elemental features of corporate law and the corporate form. In particular, distributed ledgers and blockchain technology may augment the governance capacities of shareholders, while artificial intelligence may reduce reliance on human management. How these technologies ought to be used, however, and the ends towards which they ought to be deployed, are contestable and amenable to any of a range of normative visions for the field. The core issues of corporate purpose in publicly-traded companies remain unchanged, even if they take on new aspects in an evolving technological landscape. New technologies, far from permitting us to dispense with these issues, in fact require us to return to them with renewed vigour and creativity, while acknowledging the inevitable contingency of prevailing normative views on corporate law and governance.Footnote 175
