To be clear: I like the manifesto below; I agree with it! But that agreement is not because it’s also published as a journal article. Instead, I believe it because, as a manifesto, it demands change now in terms I understand and appreciate historically.
Since my argument depends on the definition of “manifesto” I use, here’s mine:
Always layered and paradoxical, [a manifesto] comes disguised as nakedness, directness, aggression. An artwork aspiring to be a speech act—like a threat, a promise, a joke, a spell, a dare. You can’t help but thrill to language that imagines it can get something done. You also can’t help noticing the similar demands and condemnations that ring out across the decades and the centuries— something will be swept away or conjured into being, and it must happen right this moment. . .This is a form that asks readers to suspend their disbelief, and so like any piece of theater, it trades on its own vulnerability, invites our complicity, as if only the quality of our attention protects it from reality’s brutal puncture. A manifesto is a public declaration of intent, a laying out of the writer’s views (shared, it’s implied, by at least some vanguard “we”) on how things are and how they should be altered. Once the province of institutional authority, decrees from church or state, the manifesto later flowered as a mode of presumption and dissent. You assume the writer stands outside the halls of power (or else, occasionally, chooses to pose and speak from there). Today the US government, for example, does not issue manifestos, lest it sound both hectoring and weak. The manifesto is inherently quixotic—spoiling for a fight it’s unlikely to win, insisting on an outcome it lacks the authority to ensure.
In 2023, Mark Duffield and Nicholas Stockton published, “How capitalism is destroying the Horn of Africa: sheep and the crises in Somalia and Sudan,” in the peer-reviewed Review of African Political Economy (https://www.tandfonline.com/doi/full/10.1080/03056244.2023.2264679). To its credit, the Review publishes “radical analyses of trends, issues and social processes in Africa, adopting a broadly materialist interpretation of change.” And what a breadth of fresh air this article is.
First, its tone is direct, its language unequivocally materialist in the great manner of yesterday, its focus is on the marginalized, and, equally important, how could we not want change after reading this?
We present in outline an historically and empirically grounded explanation for the post-colonial destruction of the nation states of Somalia and Sudan. This is combined with a forecast that the political de-development of the Horn, and of the Sahel more generally, is spreading south into East and Central Africa as capitalism’s food frontier, in the form of a moving lawless zone of resource extraction. It is destroying livelihoods and exhausting nature. Our starting point is Marx’s argument that the historical growth and the continuing development of capitalism is facilitated through what he called ‘primitive accumulation’. With regard to the current situation in the Horn, there is a sorry historical resonance with the violent proto-capitalist land clearances that took place from the sixteenth century onward in England, Ireland and Scotland and then in North America. While today, as in Darfur, this may be classified as genocide, the principal purpose of land clearances is to convert socially tilled soils and water resources used for autonomous subsistence into pastures for intensive commercial livestock production, which now in Somalia and Sudan amounts to nothing short of ‘ecological strip mining’.
To repeat, how could you (we) not want radical change when reading further:
. . .we argue that the trade is intimately connected with the deepening social, economic and ecological crisis of agro-pastoralism in the region and the way that livestock value is now realised. Underlying the empirical data is the intensification of an environmentally destructive mode of militarised livestock production that, primarily involving sheep, is necessarily expansive, land-hungry, livelihood destroying and population displacing. Sustained by raw violence and strengthened by United Arab Emirates (UAE) and Saudi investment in Red Sea port infrastructure, the Horn and the Gulf are locked into a deadly destruction–consumption embrace.
More, there is a singular cause and it is clear: “This internationally facilitated mode of appropriation, with its associated acts of land clearance, dispossession and displacement, is the root cause of the current crisis.” Nor is there anything really complex about this:
The depth and cruel nature of the changes in Sudan and Somalia’s agro-pastoral economies cannot reasonably be attributed only to environmental change, scarcity-based inter-ethnic conflict, or avaricious generals per se. To lend these arguments weight, some hold that they combine to produce a ‘complex’ emergency. The only complexity, however, is the contortions necessary to fashion a parallel universe that usefully conceals the rapacity of capitalism. Particularly cynical is the claim, for example, that Somalia’s long-history of de-development is the result of climate-change-induced drought. It is no accident that the same international powers and agencies fronting this claim have, for decades, been active players in the Horn’s de-development.
You cannot imagine how much I want to believe these words! And I take that to be a good measure of just how effective a manifesto this manifesto is, at least for someone like myself.
Manifestos are their own public genre, whatever the publication venue. This is not a policy memo whose second sentence after the problem statement is the answer to: What’s to be done and how? But then we would never look to manifestos for the devil in the details, would we?
Keep the oil in the ground; stop fossil fuels; just don’t emit; just stop cutting down trees; just stop having babies; just stop using plastics; stop defense spending; stop being imperialist; stop techno-solutionism; dismantle capitalism; transform cities; save biodiversity; never forget class, gender, race, inequality, religion, bad faith, identity. . . and. . .
A huge category mistake is committed when conflating (1) the unfolding and interrelated consequences on life, property and markets of, say, a hazardous liquids pipeline explosion on populations and property and (2) the explosion’s consequences for the interconnected critical infrastructure system (ICIS) for those hazardous liquids, which includes not just these pipelines and associated refineries, but also–just as significantly–the electricity and water infrastructures that the former depend upon in real time.
So what? To equate “the system” with the impacts of the spread and interaction of knock-on population-and-property consequences of failure (Cf) is to identify the chief problem as one the lack of systemwide management of Cf. It is as if many official units (jurisdictional, administrative) were not or are not doing their job.
Yet the ICIS is in fact manage in real time by the control rooms of the respective infrastructures (which in turn are regulated systemwide by fewer regulators of record). That is, “coordination” can be taking place within the ICIS around shared overlapping system control variables, albeit not (or to a lesser) extent in the “system” of interconnected impacts (Cfs) from the explosion.
People seeking to control uncertain task environments pay a “control premium,” like the poverty premium (where poor people have to pay more for key services, such as insurance, credit, energy, shelter). An exclusive focus on control strategies cost them—and us—more than would be the case were they able to manage in the face of uncertainty. It’s like they are demanding money with menaces from us.
II
Here’s a different analogy to reinforce the point.
Compare algorithmic decisionmaking (ADM) and the current technology for gene editing known by the acronym, CRISPR. When it comes to ADM, the worry is that we don’t know how the algorithm works. What’s happening, we ask, because of the cultural biases imported via the original data into the algorithm? As for CRISPR, the worry is that, even when we know that this rather that gene is being edited, we’re still not sure it’s the right thing to do.
Suppose we had a CRISPR for ADM: We could go into the algorithm and excise cultural bias. But even then you’d have to worry about, e.g., what is bias to some is not to others. For that matter, is there any doubt that a new mechanism promising greater control in addressing one worry won’t produce another worry, equally if not more important?
The upshot: Control cannot answer the questions control poses.
III
So what?
It’s hard to believe, by way of answer, that all the talk about artificial intelligence (AI) “controlling” behavior will not need to be far more nuanced and contextualized for policy and management.
Consider the already-highly automated underwater oil and gas exploration. Alarms produced by autonomous systems during turbulent seas have turned out to be false alarms occurring. Indeed, and this is the point, operating at a higher level of autonomy and having to cope with recurring false alarms may no longer permit the real-time operators to revert, just-in-time, to lower levels of autonomy, e.g., managing (not “controlling”) via more manual operations, as and when nothing else works.
“I have never asked that all trees have one bark” Gotthold Ephraim Lessing, Nathan the Wise
I
Each angel is its own species, Thomas Aquinas tells us. “Why mightn’t there be, somehow, a new science for every object?,” asks Roland Barthes. In those I see the rightness and certainty of context I also find in the lines of A.R. Ammons:
though I have not been here long, I can look up at the sky at night and tell how things are likely to go for the next hundred million years: the universe will probably not find a way to vanish nor I in all that time reappear.
Why not each its own science and species, having-been for the rest of eternity?
II
Is our sense of incompleteness the felt part of an irreducible particularity of each being, that sense we never body forth as representative or total? This sense has been professed by very different world systems and believers:
“The [French] Constitution of 1795, like its predecessors, was made for man. But there is no such thing as man in the world. In my lifetime I have seen Frenchmen, Italians, Russians, etc.; thanks to Montesquieu, I even know that one can be Persian. But as for man, I declare that I have never in my life met him; if he exists he is unknown to me,” declared conservative critic, Joseph de Maistre.
Or consider the more recent lines of poet, Fernando Pessoa,
They spoke to me of people, and of humanity. But I’ve never seen people, or humanity. I’ve seen various people, astonishingly dissimilar, Each separated from the next by an unpeopled space
Imagine you are a new graduate student assigned to undertake a comprehensive survey of the many narrative analytical approaches to public policy and management. All of this is fresh for you and, naturally, you look first for current literature reviews of primary approaches. What would be among your first findings?
By way of answer, a major review of 25 years of discourse analyses (DAs) around policymaking (Leipold, Feindt, Winkel & Keller 2019, p. 448) notes:
Although its name resembles Roe’s (1994) Narrative Policy Analysis, [the Narrative Policy Framework, NPF] is not part of the landscape of discourse analysis covered in this special issue. NPF scholars clearly distance their work from post-structural ontologies and methods and position NPF as a positivist alternative to the study of policy discourses and narratives. Their aim is to provide an analytical basis for an ‘empirical’ access to socio-cultural meaning structures (i.e. generated through the analysis of large sets of text data) that stands alongside established ‘positivist’ frameworks for the analysis of the policy process (for an overview of these approaches see Weible & Sabatier, 2017). . . .
with the review authors adding that “discourse analytical approaches were ‘ignored’ in political science handbooks ‘largely on grounds that they don’t follow scientific norms’ of ‘clarity, hypothesis-testing, acknowledgement of uncertainty, etc.’ (Sabatier & Weible, 2014, p. 11).” As the passage more than hints, the same exclusionary focus of discourse analysis to NPF is found in the NPF literature with respect to DAs.
Take, by way of example, a major review of decades’ work on NPF (Schlaufer, Kuenzler, Jones & Shanahan 2022). Both the latter review and the Leipold et al (2019) review call for more interdisciplinary, cross-context research. But our graduate student quickly notes that the NPF review does not pursue the facts:
That discourse analysis has also analyzed policy narratives in the same areas as NPF, notably the environment;
That policy processes, a core focus of NPF, have also been a core focus of discourse analysis: “How does discourse analysis contribute to our understanding of environmental policy processes. . .?” (Leipold et al 20l9, p. 446);
That a central place given in NPF to heroes, villains, victims and plot in policy narratives has also been a centerpiece of other narrative approaches, notably the narrative-networks model (NNM) of political scientist Helen Ingram and her colleagues, Raul Lejano and Mrill Ingram (e.g., Ingram, Ingram & Lejano 2019); and
That mixed methods and interviews, found to be prominent in the NPF review, are equally prominent in other narrative analytical approaches. To take one from many examples, Q-methodology is used in the NPA work of van Eeten (2006), discourse analysis work of Stevenson (2019), and the NPF work of Brumbaugh & Rupp (2020). Q-methodology and interpretative policy analysis (IPA) have had a long association as well (e.g., Lundberg 2020).
Such studied disregard, it must be stressed, has been mutual and reciprocated. Where the discourse analysis review mentions the Weible and Sabatier framework of processes, it lists by name only allied discourse approaches; the NFP review lists by name only the Weible and Sabatier processes, seeing no need to recognize others, however well-established.
So what?
The reader might now expect me to argue and make the case for cross-fertilization, or at least better interchange, between and among the narrative analytical approaches. The aim here is more modest. An important step has to be taken before pinpointing areas of mutual benefit. That is, attempts at cross-fertilization are premature where the methodological reasons that render narrative analytic approaches different are not first better identified and appreciated.
These differences go well beyond conventional binaries of positivist versus post-positivist, quantitative versus qualitative, holistic versus reductionistic, incommensurable versus commensurable, and scientific versus interpretivist. Even where the terms have been used by founders and acolytes of the different narrative analytical approaches, the binaries have been complicated from the very get-go.
Qualitative versus qualitative—and yet the first example of formal quantitative network analysis among the approaches discussed here is the irrigation case study of Narrative Policy Analysis (1994). Positivist versus post-positivist—and yet you can imagine the perplexity of those in other narrative analytical traditions, grounded as they are in case analysis, to the statement, “The NPF brought empiricism to the study of narratives” (Schlaufer et al 2022, p. 261). Scientific versus interpretivist—and yet major semioticians (notably, A.J. Greimas [1987, pp. xl – xli, 121ff]) would have treated as risible any intimation that theirs wasn’t a rigorous, scientific project.
The argument is that, in light of these and like complications, there needs to be greater clarity about three under-acknowledged methodological issues prior to comparing these narrative analytical approaches and drawing lessons from and for each other. Three deserve special mention.
Different approaches ask different questions.
Each of the respective narrative approaches proffers different interrogations of policy narratives. This would be banal, were it not for the logical and empirical implications.
Perhaps the best example is NPA when compared to both NPF and DAs: the former asks how narratives underwrite and stabilize the assumptions of decision making under uncertain and complex circumstances; the latter two approaches make no such claim. For them, as for other approaches, narratives are stories with various uses, including—and certainly not excluding—that of persuading others of what is already taken to be known or true.
Being clearer about key-question differences would aid in identifying areas of useful cross-approach interchange. For example, the NPF review notes: “However, not much is known about how these narratives affect the policy process or policy outputs. That is, where policy narratives originate, whom they impact, and to what effect are all important policy narrative questions that are rarely addressed simultaneously” (Schlaufer et al 2022, p. 259). That latter point, nevertheless, is a settled matter from the perspective of some narrative approaches. It’s precisely how narratives originate, whom they impact, and what are their other effects that are demonstrated in the environmental governance focus of NNM.
Also, NPA was extended to demonstrate how infrastructure operators and emergency managers underwrite and stabilize real-time decision making under turbulent conditions. Systemwide failure scenarios—a type of policy-relevant narrative—affect real-time decisions, even if (and especially because) some contingencies cannot be narrativized at that point in decision making. More, the well-documented real-time use of failure scenarios complicates any kind of micro-meso-macro distinction (a key feature of the NPF, for example). Stabilization, even in normal operations in large critical infrastructures, takes place frequently in the form of team situation awareness of systemwide operations interconnected with the operations of other critical infrastructures (Roe & Schulman 2008, 2016).
“Meso-levels” may become clear only later on, though that too can’t be assumed beforehand. Nor is a persisting lack of clarity over levels-of-analysis unfamiliar in the social sciences. “Interdependence” has long been notorious for the absence of agreed-upon empirical measures (see La Porte 1975/2015).
Methodological advances in the approaches are not cabined to scholarly confines.
Reviews of the narrative analytical approaches are not as comprehensive as they might first appear. This happens especially when the gray literatures of practicing policy analysts, public managers and scholar-activists are not canvassed. The reasons are various, including but not limited to the case material not entailing a focus, e.g., on discourse coalitions as in many DAs (Hajer 1993), the more-than-human in the narrative networks of NNM (Ingram et al 2019), the four-step method of an NPA (Roe 1994), or the specific assumptions of NPF (Schlaufer et al 2022).
Consider a timely example. A recent working paper, Understanding the role of narratives in humanitarian policy change and published by the London-based Overseas Development Institute (Saez & Bryant 2023), is explicitly produced for a non-scholarly audience and states it is grounded in NPF. Yet it is not clear how the working paper’s highly original and useful Figure 1 (Ibid, p. 16) derives from the NPF (note too the wider appeal to “discourses” at the figure’s center):
Arguably more important, Figure 1 highlights the fact that policy narratives can be geared around delaying rather than supporting this or that policy and management (for other examples, see Hermann 2024). In other words, different narrative analytical approaches must be assumed to have different granularities, irrespective of (in spite of) any prejudgments about “levels of analysis.”
The differences in granularities open up all manner of possibilities for recasting (think: renarrativizing) so-called wicked policy problems (Roe 2023). For example, while DAs from the get-go focused on “the policy process,” their more recent developments differentiate by way of specifics between and among “maneuvering within a given discursive framework, navigating between different and conflicting discourses, or transforming existing discourses” (Lynggaard & Triantafillou 2023, p. 1).
The methodological challenge of evaluating different types of policy narratives.
Although ubiquitous, the term, “policy narratives,” is not helpful unless the narratives in question are first differentiated. We just saw the methodological need to distinguish and take note of the more granular uses of failure scenarios in large socio-technical systems. It is quite another matter to single out policy narratives that are conspiracy theories which various leaders, past and present, believe and act upon in their advocacy and sometimes policy-making.
The basic question here is the perennial one of how to evaluate different types of policy narratives, or more specifically: What are the evaluative criteria, if any, of the different narrative analytic approaches? Suffice to say, there are approaches that rely on the policy processes they study or assume for making any such evaluation. Other approaches have evaluative criteria independent of those processes or profess none at all.
An example helps. A complete policy narrative in NPA meets several criteria, including: the narrative—its story with beginning, middle and end, or argument with premises and conclusions—is one that takes seriously that the policy or management issue is complex, uncertain, and/or conflicted. One chief assumption of conspiracy theories is that nothing happens by accident (Barkun 2013). From the NPA perspective, such narratives are inferior when compared to other narratives that posit accidents, happenstance, chance and luck happening in and through policy processes for the same issues.
This is not an argument for one approach over another—rather, the point again is methodological. In particular, it might be far better to think of evaluative criteria as a major link between policy narratives and the policy processes, including the latter’s technologies and actionable contingencies, in which the narratives are embedded.
Upshot.
A mentor, political scientist Aaron Wildavsky, famously wondered: If planning is everything, maybe it is nothing. So too all of us might want to ask of those policy narratives said to be everywhere and important for almost everything. Are we overthinking narratives (see Brooks 2022)?
As the start to an answer, a hypothesis is proposed: A less-recognized but major consequence of the siloed approaches to the study of policy narratives has been to slight and confuse the role of very real, very messy policy processes in determining and differentiating types of and criteria for narratives (Roe 2013).
Sources
Barkun, M. (2013). A Culture of Conspiracy: Apocalyptic Visions in Contemporary America. Second Edition, Berkeley, CA: California University Press.
Brooks, P. (2022). Seduced by Story: The Use and Abuse of Narrative. The New York Review of Books: New York, NY.
Brumbaugh, A. and J.H. Rupp (2020). Wabash CarbonSAFE. Subtask 4.1 – Application of Policy Frameworks for Improved Carbon Capture and Storage. Social Site Characterization & Stakeholder Engagement. Report Number: DOE/ FE0031626‐2. U.S. DOE Cooperative Agreement Number: DE-FE0031625, Champaign, IL.
Greimas, A.J. (1987). Qn Meaning Selected Writings in Semiotic Theory. Trans. P. J. Perron and F. H. Collins, Theory and History of Literature, Volume 38, Minneapolis MN: University of Minnesota Press.
Hajer, M. (1993). Discourse coalitions and the institutionalization of practice: The case of acid rain in Britain. In F. Fischer & J. Forester (Eds.), The Argumentative Turn in Policy Analysis and Planning (pp. 43–76). Durham, NC: Duke University Press.
Ingram, M., H. Ingram and R. Lejano (2019). Environmental Action in the Anthropocene. Journal of Environmental Policy and Planning 21(5): 492–503.
La Porte, T. R. (Ed.). (1975/2015). Organized Social Complexity: Challenge to politics and policy. Princeton, NJ: Princeton University Press.
Lejano, R., M. Ingram, and H. Ingram (2013). The Power of Narrative in Environmental Networks. Cambridge, MA: MIT Press.
Leipold, S. P. Feindt, G. Winkel & R. Keller (2019) Discourse analysis of environmental policy revisited: traditions, trends, perspectives, Journal of Environmental Policy & Planning, 21:5, 445-463
Lundberg, A. (2020). Viewpoints about Educational Language Policies: Multilingualism in Sweden and Switzerland. Malmö Studies in Educational Sciences No. 90, Faculty of Education and Society, Malmö University, Sweden.
Lynggaard, K and P. Triantafillou (2023). Discourse analysis and strategic policy advice: manoeuvring, navigating, and transforming policy. Journal of European Public Policy, DOI: 10.1080/13501763.2023.2217846
Roe, E. (1994). Narrative Policy Analysis: Theory and Practice. Durham, North Carolina: Duke University Press.
——- (2013). Making the Most of Mess: Reliability and policy in today’s management challenges. Durham, NC: Duke University Press.
——- (2023). When Complex is as Simple as it Gets: Guide for Recasting Policy and Management in the Anthropocene. IDS Working Paper 589, Brighton: Institute of Development Studies, DOI: 10.19088/IDS.2023.025
Roe, E. and P.R. Schulman (2008) High Reliability Management, Stanford CA: Stanford University Press.
——- (2016). Reliability and Risk, Stanford CA: Stanford University Press.
Schlaufer, C., J. Kuenzler, M.D. Jones, & E. Shanahan (2022). The Narrative Policy Framework: A Traveler’s Guide to Policy Stories. Politsche Vierteljahresschrift 63:249–273.
van Eeten, M.J.G. (2001). Recasting Intractable Policy Issues: The Wider Implications of the Netherlands Civil Aviation Controversy, Journal of Policy Analysis and Management 20(3): 391 – 414. 2001.
Weible, C.M., & Sabatier, P.A. eds. (2017). Theories of the Policy Process (4th ed.). Boulder, CO: Westview.
Apologies in advance for the length, but the quoted extract is highly illuminating!
Recently, I came across the work of a graduate student, whose identity is charitable to keep hidden. The text was written rather well and included a critical review of the experimental studies conducted on one of my research topics: therefore, my name and the name of the colleague who worked with me on that line of research, Marco Marini, often appeared in the text, and consequently also in the final bibliography. The consultation of the references, however, had in store a few surprises. Among other entries, the following were recorded:
–Marini, M. (2013). When it’s better to choose the one you love: The effect of attractiveness biases in consumer choices. Judgment and Decision Making, 8(5), 476-485.
–Marini, M. (2019). How to get people to take risks? A choice-based measure of risk preference. PloS One, 14(1), e0209983. doi: https://doi.org/https://doi. org/10.1371/journal.pone.0209983
–Marini, M. (2019). Luring to a suboptimal option: The effect of payoff reduc- tion in a risky choice framing. Judgment and Decision making, 14(2), 198- 207.
–Marini, M. (2020). The asymmetrically dominated compromise effect in a dynamic setting. Journal of Economic Psychology, 76, 102-257.
–Paglieri, F. (2009). The attractiveness of decoys in economic contexts: An experimental investigation. Judgment and Decision Making, 4(4), 335-342.
Formally, this bibliography extract is flawless: the entries are correctly formatted according to the standards of the American Psychological Association (APA), the relevant information is all present, the articles are consistent with the topic of the student’s assignment, and the titles of the various contributions are, objectively, quite intriguing. The only problem is that… none of these publications exist!
The incident was neither a brave, subversive act of provocation (to demonstrate that university instructors no longer read carefully the written assignments of their students), nor a symptom of terminal stupidity in the student (only a very dumb cheater would try to falsify the references of the very same people tasked with evaluating their work): instead, it was the outcome of a naïve and inappropriate use of generative AI. The student, after writing the assignment themselves and inserting the appropriate references in the text, using the author-date APA standard, had incautiously asked ChatGPT to prepare the reference list, giving it their own text as part of the prompt. Unfortunately, the software compiled a bibliographic list in full compliance with APA standards, but without any attention to the truthfulness of the information included therein.
Here, however, we are not interested in the student’s misadventures, but rather in how ChatGPT produced its output, which was certainly not random: there is method to this madness. Firstly, the journals in which the fake contributions would have appeared are plausible, both thematically, and because Marini and I have already published in those venues in the past, or in very similar ones. Secondly, the volume numbers that are mentioned refer to issues that have indeed been released, and usually the numbering and year of publication match; in one case, the entire reference (PloS One, 14(1), e0209983. doi: https://doi.org/10.1371/journal.pone.02099 83) refers to an existing article, except that it is a study on a completely different topic, i.e. gender barriers in research at the South Pole (Nash, M., Nielsen, H., Shaw, J., King, M., Lea, M.-A., & Bax, N (2019), “Antarctica just has this hero factor…”: Gendered barriers to Australian Antarctic research and remote fieldwork).
The inconsistencies that emerge upon closer inspection are also revealing: the 2020 article attributed to Marini is listed as appearing between page 102 and page 257, except that there never was a single 155-page long contribution published in that particular journal, and probably not even in others, at least in the field of economic psychology; delving deeper, one discovers that the Journal of Economic Psychology, from 2020 onwards, no longer reports the page numbers of individual articles, but only their identification number, which is composed of a 6-digit code starting with 102, and the code 102257 (that ChatGPT creatively transformed into page numbers, 102–257) corresponds to the editorial of the issue following the one cited in the invented bibliographic reference.
At other times, the system falls prey to ambiguities of meaning: the decoy effect, which was the main focus of the student’s paper, is also referred to as the attraction effect in the literature, and the word “attraction” evokes the semantic field of affects, which instead has nothing to do with the technical phenomenon in question (i.e., a shift of preferences towards an option that is manifestly superior to another inserted ad hoc, called decoy). It is because of this semantic ambiguity that ChatGPT came up with a title like “When it’s better to choose the one you love: The effect of attractiveness biases in consumer choices” – a wonderful title, by the way, which I will certainly use, as soon as the opportunity presents itself.
In short, this false output is not due to anomalies or errors in the functioning of the software, but on the contrary it illustrates perfectly what ChatGPT is built to do (and does very well): generate linguistic strings (in this case, bibliographic entries) that have the maximum probability of satisfying the user’s request, based on similar instances present in the (huge) database to which the program had access during training. What ChatGPT does not do, and cannot do due to the way it functions (at least for the time being), is consulting the real world or an internal representation of it: the system does not work by checking the state of the world and describing it, but rather by constructing responses that are maximally consistent with the vast mass of linguistic data at its disposal, whose adherence to reality is by no means guaranteed.
More than a year ago a joint statement was issued by the Center for AI Safety. It was the one sentence quoted above. Famously, it was signed by more than 350 AI experts and public figures.
Now, of course, we cannot dismiss the actual and potential harms of artificial intelligence.
But, just as clearly, these 350 people must be among the last people on Earth you’d turn to for pandemic and nuclear war scenarios of sufficient granularity against which to appraise their AI crisis scenarios.
The more you aim for just-enough reliability, the more specific and narrow are the criteria of “just how good is just-good-enough” (“you must respond within x minutes of a call. . .”). Goal displacement is risked where, e.g., meeting government regulatory compliance equates to ensuring the infrastructure’s continuous and safe provision of its critical service. In so doing, safe and continuous service reliability eventually falls to the side when system operators feel compelled to be fast enough with just enough, knowing this is never enough all the time.
If we want anything more by way of highly reliable service provision, then that is left to us, not so much as consumers or citizens, but as amateurs who are now be their own reliability managers. So goes the economists’ wet dream of just-enough reliability. Gone are the days when anyone felt comfortable with discussions that include, “Elementary economics demonstrates that. . .”
Start with what passes today as a cliche: “Policymakers also need to worry about those other factors—societal, political, economic, historical, cultural, geographical, governmental, psychological, technological, ethical, religious etc—that are so undeniably part of policy analysis and management.”
That pincushion of “etcetera’s” Indicates things are just critical enough to get a nod, but not that critical to actually name. So too Samuel Taylor Coleridge argued “matter” was treated like a pincushion whose surface was hidden by all the sensations, thoughts and properties stabbed into it.
You ask today’s version of, “What are the important factors?,” and you get a pincushion of responses affixed with all manner of “etc’s” “Hail, Muse! Et Cetera,” the poet, Byron, put it in the third canto of Don Juan.
And yet, writes Wittgenstein: “Again and again, my ‘etc’ has a limit.”
WHEN COMPLEX IS AS SIMPLE AS IT GETS 