Francis Bacon is generally known as the man who developed the scientific methodFootnote 1 in the sixteenth century, but he is also credited with coining the aphorism that “knowledge is power” (Meditationes Sacrae, 1597). As economies and society in general have grown more complex, the truth of this aphorism has only increased. In 1970, George Akerlof published “The Market for Lemons,”Footnote 2 his seminal paper on information asymmetries and how access to information can define markets and establish winners and losers. Although Akerlof focused on the used car market, his insights applied across sectors and industries. In 2001, Akerlof, Michael Spence and Joseph Stiglitz were awardedFootnote 3 the Nobel Prize in Economics “for their analyses of markets with asymmetric information.”
The insight that knowledge, resulting from having access to (privileged) information or data, is power is more relevant today than ever before. The data age has redefined the very notion of knowledge and information (as well as power), leading to a greater reliance on dispersed and decentralized datasets as well as to new forms of innovation and learning, such as artificial intelligence (AI) and machine learning (ML). As Thomas Piketty (among others) has shown, we live in an increasingly stratified world, and our society’s socio-economic asymmetriesFootnote 4 are often grafted onto data and information asymmetries. As we have documented elsewhere,Footnote 5 data access is fundamentally linked to economic opportunity, improved governance, better science and citizen empowerment. The need to address data and information asymmetries—and their resulting inequalities of political and economic power—is therefore emerging as among the most urgent ethical challenges of our era, yet often not recognized as such.
Even as awareness grows of this imperative, society and policymakers lag in their understanding of the underlying issue. Just what are data asymmetries? How do they emerge, and what form do they take? And how do data asymmetries accelerate information and other asymmetries? What forces and power structures perpetuate or deepen these asymmetries, and vice versa? I argue that it is a mistake to treat this problem as homogenous. In what follows, I suggest the beginning of a taxonomy of asymmetries. Although closely related, each one emerges from a different set of contingencies, and each is likely to require different policy remedies. The focus of this short essay is to start outlining these different types of asymmetries. Further research could deepen and expand the proposed taxonomy as well help define solutions that are contextually appropriate and fit for purpose.
1 Data asymmetries
Data asymmetries are the classic—and most commonly recognized—form of asymmetry in today’s digital marketplaces. They occur whenever there exists a divide or disparity in control of and access to data.Footnote 6 The nature of this divide can take many forms, however, depending on the relationship between data holders (“owners”) and users.
Consumer-to-business (C2B) data asymmetries, resulting from the increased generation or aggregation of consumer data by business, dominate much current public discussion. Such asymmetries have grown common with the datafication of consumption patterns and typically occur when companies collect data on their users while providing services or selling goods. For example, companies might collect data related to transaction or browsing histories or a variety of socio-demographic markers. As a result, companies often possess a disproportionate amount of data on their users— data that users may not even be aware of having surrendered. This data allows companies to sell or target advertisements,Footnote 7 optimize internal operations, train AI systems, and pursue other objectives to increase or generate market power (as a result of so-called two-sided marketplaces).Footnote 8 This increased “dataveillance” prompts ethical discussions around “data sovereignty” or “digital self determination” and how to provide back a certain level of agency to individuals about how others re-use their data as well as a search for new policy solutions.
In addition to C2B data asymmetries, various other forms of asymmetry also exist. Among the most consequential are business-to-business (B2B) asymmetries. Recent years have witnessed the emergence of a number of large data monopolies or platforms that dominate their sectors and the broader economy. These companies have access to huge amounts of data collected, processed and aggregated across various domains (such as search data, location and mobile phone data, consumer spending data, ride sharesFootnote 9) and their ability to combine and derive insights from this data or train ML algorithms results in de facto barriers to entry. There are concerns that B2B data asymmetries may be stifling innovation and competition as well as hurting the rights of consumers, leading to calls for greater regulation and better enforcement of antitrust law,Footnote 10 perhaps extending so far as to the breakup of some of these large players.
Other data asymmetries worth considering relate to business-to-government (B2G), in which government may be hampered in developing data-driven or evidence-based policies or providing targeted services without access to data and insights that the private sector may possess—a topic considered by the High Level Expert Group to the European Commission on B2G Data SharingFootnote 11 (of which I am a member). Other data asymmetries relate to government-to-society (G2S), in which data collected by the government is siloed and hoarded without transparency or without making it accessible to society at large through, for instance, open data platforms.Footnote 12
Concerns for privacy harms and increased surveillance in all of the above cases are real and require careful consideration and mitigation. Yet too often, the spectre of harm to privacy and civil liberties is used to justify limiting access to data that could provide for transformative public value. Limiting access may, however, increase the asymmetries without fully addressing the privacy concerns or establishing trust in how data is being collected and handled.
2 Information asymmetries
Data asymmetries occur when there are disparities in access to data. But even when such disparities are overcome, there often exist pervasive inequalities in the extent to which individuals and groups can actually benefit—for example, by deriving new insights or informing innovation—from their formal access. In short, stakeholders differ in their abilities to translate data into actionable information, generating information asymmetries.
The data ecology is highly complex and rapidly growing in complexity. The legibility of data depends significantly on the technical, financial, and human resources of organizations that collect, store, and access data. Smaller firms and organizations, as well as individuals, may be at a particular disadvantage, as they frequently lack the know-how to surface the signal in noisy data. Smaller organizations may similarly be at a disadvantage at a time when so much of the data ecology depends on collaboration and data integration. Their relative lack of resources may not only limit their ability to derive insights from data but also to operate as equal partners when defining terms and other parameters of collaboration.
Moreover, establishing information through, for instance, machine learning has become an increasingly computational intensive activity. For instance, the computations required for deep learning research have been doubling every few months, resulting in an estimated 300,000× increase from 2012 to 2018.Footnote 13 As such, those who have access to high computing power have an asymmetric gain to those who haven’t.
3 Intelligence asymmetries
The datafication of life and the economy has spurred the development of (and reliance upon) a number of new technologies, among the most prominent of these being ML and AI. While offering tremendous potential, these technologies also pose risks and are leading to new forms of asymmetry. I call these intelligence asymmetries.
Intelligence asymmetries occur when there is a discrepancy among actors in their ability to understand (i.e., “look under the hood of”) the algorithms and processes that are responsible for an increasingly broad range of automated decisions. Today, AI plays a role in determining the outcome of loan applications, bail and parole requests, and other important matters. How these decisions are made is increasingly opaque and contested.Footnote 14 The fact that different stakeholders have differential capacity and insight into both the algorithmic processesFootnote 15 themselves and the underlying dataFootnote 16 (used to derive those algorithms through ML) is an increasingly worrying form of asymmetry. It is now widely recognized that AI algorithms often contain inherent biasesFootnote 17; addressing intelligence asymmetries through greater traceability is therefore often a matter of social justice and wider socio-economic equity.
Intelligence asymmetries are a growing problem in the data ecosystem and, if left unaddressed, can undermine trust in those organizations and initiatives that are focused on re-using data for other purposes than collected (such as Sidewalk Labs in Toronto).Footnote 18 Overall, there is a need for new norms and standards regarding how training data is collected and used, and a general insistence on greater transparency (and explainability) when it comes to AI-driven decision-making.
The above has outlined three types of inter-related asymmetries that increasingly define the data age. As I show, these asymmetries are important not only to achieve greater equity in access to data but also because they are at the root of many pernicious and growing socio-economic power inequalities. This essay has been primarily descriptive. While I have described the problem, we need a lot more exploration and experimentation to design adequate solutions.
Nonetheless, there does exist an emerging set of possible solutions that could provide a toolbox, and these possibilities should be explored more fully in future research. For example, some argue for a greater focus by policymakers and others on data liquidity (and portability),Footnote 19 which would enhance citizens’ and others’ agency and determination over data. This focus could also generate an ecosystem of responsible data exchanges. Data holders and demand side actors could also experiment with new and emerging operational models and governance frameworks for purpose-driven, cross-sector data collaborativesFootnote 20 that bring to bear previously siloed datasets. In addition, there is a need to further define and professionalize the notion of “data stewards”,Footnote 21 individuals or teams who are tasked with managing and providing access to data and its ethical and responsible re-use within organizations. The potential of emerging technologies such as Distributed Ledger Technologies and others to address information asymmetriesFootnote 22 also need further attention.
Beyond these specific possibilities, there is a need to demystify data, provide transparencyFootnote 23 into existing relationships—how digital systems collect and process data, their intended purpose and who is responsible for that data activity- and crucially, provide mechanisms for feedback. No citizen or stakeholder is removed from the ramifications of datafication, including its resulting asymmetries. Francis Bacon was right all along: Knowledge is power, and the extent to which organizations can address these asymmetries will depend in large part on public engagement (such as data assemblies),Footnote 24 which, in turn, relies on greater literacy and awareness on the part of citizens and policymakers.
The above are small steps in the direction of raising public awareness, and thus empowering citizens with knowledge about the data age and the need for improved data governance. Much more needs to be done to address the ethical and concrete ramifications of data, information and intelligence asymmetries.
“Scientific method,” Wikipedia, last modified July 22, 2021, https://en.wikipedia.org/wiki/Scientific_method.
“Popular information,” The Nobel Prize, https://www.nobelprize.org/prizes/economic-sciences/2001/popular-information/.
Open Data’s Impact, https://odimpact.org/.
Leigh Dodds, “What is data asymmetry?,” March 24, 2017, https://blog.ldodds.com/2017/03/24/what-is-data-asymmetry/.
Kirpalani and Philippon (2020).
Algorithmic Labor and Information Asymmetries: A Case Study of Uber’s Drivers, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2686227
High-Level Expert Group on Business-to-Government Data Sharing, “Towards a European strategy on business-to-government data sharing for the public interest,” https://www.euractiv.com/wp-content/uploads/sites/2/2020/02/B2GDataSharingExpertGroupReport-1.pdf.
Open Data Policy Lab, https://opendatapolicylab.org/.
Schwartz et al. (2020).
Julia Angwin, Jeff Larson, Surya Mattu and Lauren Kirchner, “Machine Bias,” ProPublica, May 23, 2016, https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing.
Andrew Lee Park, “Injustice Ex Machina: Predictive Algorithms in Criminal Sentencing,” UCLA Law Review, February 19, 2019, https://www.uclalawreview.org/injustice-ex-machina-predictive-algorithms-in-criminal-sentencing/.
Buolamwini and Gebru (2018).
Joan Palmiter Bajorek, “Voice Recognition Still Has Significant Race and Gender Biases,” Harvard Business Review, May 10, 2019, https://hbr.org/2019/05/voice-recognition-still-has-significant-race-and-gender-biases; William Crumpler, “The Problem of Bias in Facial Recognition,” Center for Strategic and International Studies, May 1, 2020, https://www.csis.org/blogs/technology-policy-blog/problem-bias-facial-recognition
Flynn and Valverde (2019).
Juliet McMurren and Stefaan Verhulst, “Data to Go: The Value of Data Portability as a Means to Data Liquidity,” October 26, 2020, https://medium.com/data-stewards-network/data-to-go-the-value-of-data-portability-as-a-means-to-data-liquidity-682bb39368e0.
Data Collaboratives, https://datacollaboratives.org/.
Stefaan Verhulst, Andrew Zahuranec, Andrew Young and Michelle Winowatan, “Wanted: Data Stewards: (Re-)Defining The Roles and Responsibilities of Data Stewards for an Age of Data Collaboration,” March 2, 2020, https://medium.com/data-stewards-network/wanted-data-stewards-re-defining-the-roles-and-responsibilities-of-data-stewards-for-an-age-of-274c2d5f033a.
Stefaan Verhulst, “Information Asymmetries, Blockchain Technologies, and Social Change,” July 24, 2018, https://sverhulst.medium.com/information-asymmetries-blockchain-technologies-and-social-change-148459b5ab1a.
Digital Trust for Places and Routines, https://dtpr.helpfulplaces.com/.
The Data Assembly, https://thedataassembly.org/.
Akerlof G (1970) The market for “lemons”: quality uncertainty and the market mechanism. Q J Econ 84(3):488–500. https://doi.org/10.2307/1879431
BuolamwiniJ, Gebru T (2018) Gender shades: intersectional accuracy disparities in commercial gender classification. Proc Mach Learn Res 81: 1–15. http://proceedings.mlr.press/v81/buolamwini18a/buolamwini18a.pdf
Flynn A, Valverde M (2019) Planning on the waterfront: setting the agenda for Toronto’s ‘smart city’ Project. Plan Theory Pract 20(5):769–775. https://doi.org/10.1080/14649357.2019.1676566
Kirpalani R, Philippon T (2020) Data sharing and market power with two-sided platforms. Natl Bureau Econ Res. https://doi.org/10.3386/w28023
Piketty T (2020) Capital and ideology. Harvard University Press, Cambridge. http://piketty.pse.ens.fr/en/ideology
Schwartz R, Dodge J, Smith NA, Etzioni O (2020) Green AI. Commun ACM 63:54–63
SrinivasanD (2018) The antitrust case against Facebook. Berkeley Bus Law J 16(1). https://ssrn.com/abstract=3247362.
Zuboff S (2015) Big other: surveillance capitalism and the prospects of an information civilization. J Inf Technol 30(1):75–89. https://doi.org/10.1057/jit.2015.5
Curmudgeon Corner is a short opinionated column on trends in technology, arts, science and society, commenting on issues of concern to the research community and wider society. Whilst the drive for super-human intelligence promotes potential benefits to wider society, it also raises deep concerns of existential risk, thereby highlighting the need for an ongoing conversation between technology and society. At the core of Curmudgeon concern is the question: What is it to be human in the age of the AI machine? -Editor.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Verhulst, S.G. The ethical imperative to identify and address data and intelligence asymmetries. AI & Soc (2022). https://doi.org/10.1007/s00146-022-01410-5