By Javier Surasky
Introduction
What if
some of the most persistent problems of artificial intelligence were, at their
core, not as new as we tend to believe? Long before we spoke of opaque
algorithms, autonomous systems, or distributed responsibility, other cultures had already reflected through myths on how to govern what exceeds human capacity for control.
On this
occasion, we explore myths from Asia, particularly from China, Japan, and the
Bagobo people of the Philippines.
As in our
previous posts on Greco-Roman and Indigenous peoples of African myths, our methodology is based on a heuristic approach to myth that allows us
to put under tension elements present in current debates on AI. For us, myths
function as tools for thinking about aspects that, in their technological and
Western framing, tend to remain hidden.
In earlier
blog posts, we have already presented this methodology and its mode of
application in detail, including the consideration of myth as a form of thought
proper to its originating peoples and as situated knowledge, as well as the
safeguards we adopt to avoid falling into practices of knowledge extractivism.
We therefore refer those who wish to know more to those texts.
This time,
the myths will lead us to question elements such as accountability, algorithmic
auditing, distributed responsibility, transparency and explainability,
legitimacy, limits of use, anthropomorphism, morality, overconfidence, and bias
in AI.
Before
fully engaging in this task, we recall, together with Benasayag and Pennisi
(2024:123), their observation that in the “transition from cosmogonies in which
the gods are the metaphor of what is ungovernable for humans, where fear of
spirits or natural and supernatural entities fulfills a regulatory function, to
modernity, in which there are only humans who, in the absence of gods and
monsters, fear themselves, today one more step is taken: fear of a superhuman
technical power.”
For this
reason, our reading is directed not only at AI systems as objects of analysis,
but also at expert practices, technical, legal, and regulatory, that contribute
to instituting them as governable, legitimate, and acceptable.
AI, democracy, and accountability: the Jade Emperor
In Chinese
tradition, the structure of the cosmos is presented as an administrative order:
the divine operates as a system of government with hierarchies and distributed
functions that prevent chaos. This idea is expressed explicitly in the image of
a celestial bureaucracy organized by portfolios.
This
mythical vision invites us to think of AI as an organized system in which
roles, competences, and procedures are decisive, which introduces “noise” into
calls for greater transparency as a general way to resolve the problems posed
by AI, coinciding in this respect with Kroll et al. (2017:633): “We challenge
the dominant position in the legal literature that transparency will solve
these problems.”
Read
critically, the celestial bureaucracy represents an ideal of order linked to a
specific mode of governance, which leads us to recover Jacques Ellul’s warning
that, once a technical system is established, it tends to justify itself
through procedures that displace human responsibility: “Technique has become
autonomous; it has fashioned an omnivorous world which obeys its own laws and
which has renounced all tradition” (Ellul, 1964:79).
The
multiplication of instances, mandates, and procedures does not guarantee accountability, but may instead produce structural irresponsibility. Returning
to Ellul (1964:95): “No one is responsible for anything any longer;
responsibility is dissipated through the technical system.”
The myth of
the Jade Emperor allows us to read a risk that hangs over AI governance: its
institutional design, far from resolving the problem of responsibility,
administers it, in a dynamic that finds an echo in Foucauldian analyses of
governmentality, which describe the shift from government by law toward
security diapositives that operate through regulation, normalization, and risk
management: “Security mechanisms have the function of responding to a reality
in such a way that this response cancels out the reality to which it responds”
(Foucault, 2007:47).
This shift
implies that AI governance is oriented primarily toward keeping the system
within acceptable operating thresholds, even when its effects are
controversial. Bureaucracy, celestial or technical, sustains its authority not
because it resolves conflicts (chaos), but because it absorbs them through
management.
Seen in
this way, the myth reminds us that institutionalization itself can become a
technology for displacing responsibility, especially in complex and opaque
systems, creating an order produced, maintained, and legitimized by expert
communities that, by fragmenting responsibility in the name of technical
complexity, contribute to diffusing attribution for harms.
Transparency to preserve the status quo: Amaterasu and the mirror
In the
Shintō cycle, the stability of the world depends on the presence of the
sun-goddess Amaterasu, who, after a series of intolerable offenses, withdraws
into a cave, and as a result “the source of light disappeared, and the whole
world was plunged into darkness” (Anesaki, 2015:23). Darkness is seen as the
condition of possibility for disorder.
Faced with
the crisis, the gods deliberate: “as the result of this consultation, there
arose a number of things of divine efficacy, such as mirrors” (Anesaki,
2015:23). Finally, Amaterasu emerges from her cave, attracted by her reflection
in a mirror, and with this, the light and order are restored.
A first
intuitive reading of the myth is that, without light, the world becomes
ungovernable. In the field of AI, this intuition reappears forcefully when
relevant decisions rely on opaque systems. As Coeckelbergh (2023:103) notes,
“there is a problem of responsibility and legitimacy when decisions are made based
on AI recommendations.” The “mirror” functions here as a heuristic: restoring
legitimacy does not mean trusting again, but producing conditions of
visibility.
But the
myth also reminds us that the return of light does not equate to justice, but
rather to the restoration of an unquestioned order of governance. In
Foucauldian terms, visibility is not simply a normative value, but a technology
of government: in describing security dispositifs, the French philosopher
states that “The aim is not to eliminate phenomena, but to keep them within
acceptable limits” (Foucault, 2007:21).
Applied to
AI, this implies that making a system’s functioning visible (transparency) does
not necessarily mean opening it up to political contestation but integrating it
into an acceptable regime of risk management. In this sense, transparency can
operate as a curtain in front of political conflict: instead of opening debate
about ends and limits, it offers an image of sufficient control to sustain the
system’s continuity.
This
warning finds a direct echo in the classic critique of algorithmic
transparency. Ananny and Crawford caution that “The implicit assumption behind
calls for transparency is that seeing a phenomenon creates opportunities and
obligations to make it accountable” (Ananny & Crawford, 2016:2), when in
fact these are two different things.
The myth of
Amaterasu problematizes this assumption. The mirror shows but does not judge; it illuminates, but does not redistribute power. Light restores order by
rendering it legible, not by making it more just or by creating obligations
among administrators. The myth even warns against naive trust in explanatory
devices, as Rudin (2019:206) notes: “Explanations are often not reliable, and
can be misleading.”
Read
critically, the myth of Amaterasu results in an uncomfortable warning: making
things visible is a form of governing that does not imply change, justice, or
responsibility for the decisions of those who govern, but can instead be a way
of “maintaining order” under the pretext of confronting the chaos that its
extinction would entail.
Anthropomorphism and moral status: Pamalak
In the
Bagobo tradition (Mindanao, Philippines), the myth of Pamalak presents an
initially blurred boundary between the human and the animal. Before the
creation of humankind, tradition holds that “monkeys once behaved and looked
like humans”; and that monkeys “only acquired their current appearance when
Pamalak decided to create humankind as a separate race” (Storm, 2006:56). The
similarity between the two is not denied, but managed through the external
imposition of a decision that establishes an ontological difference between
them.
The myth
suggests that prior indeterminacy required a separation operation. When
something “appears” human, it is not enough to describe the resemblance: it
becomes necessary to decide what status is to be recognized, because from that
decision follow responsibilities, obligations, and legitimate forms of
authority.
In the
field of AI, this problem appears in anthropomorphism: “the seed of the
prevailing ontological confusion is the set of anthropomorphisms and
zoomorphisms that come bottled with AI” (Madrid Casado, 2024:121). Yet Pamalak
invites us to shift the usual reading: the boundary does not become blurred
because conceptual criteria are lacking, but because ambiguity remains
functional until overcoming it becomes more important (more functional) than
sustaining indeterminacy.
In AI, the
ease with which attributions of understanding and agency are induced in
machines had already been observed by Weizenbaum in describing the case of
ELIZA: “ELIZA shows, if nothing else, how easy it is to create and maintain the
illusion of understanding” (Weizenbaum, 1966:42), a confusion that carries
practical effects. As Reeves and Nass (1996:5) show, people tend to interact
with media technologies as if they were social actors, generating patterns of
overconfidence, which Parasuraman and Riley (1997:232) describe as a “misuse”
of process automation, emphasizing that the problem is not technical but
behavioral: delegating when one should not, and ceasing to monitor when
monitoring is necessary.
Although it
is often assumed that these problems can be addressed through better design or
greater literacy, the myth of Pamalak suggests that the “humanization” of the
non-human is a condition of possibility for delegation, insofar as it
facilitates the creation of trust, empathy, interactive fluency, and
acceptance.
But this
functionality is not neutral: it favors institutional arrangements in which the
“humanization” of the digital becomes an objective to be pursued, ultimately
helping to displace responsibility. In other words, “Artifacts have politics”
(Winner, 1986:121). Read critically, Pamalak does not invite the elimination of
resemblance, but rather the recognition of its political character.
Conclusion
This work
has examined three Asian myths as critical heuristics for interrogating
contemporary problems of AI, not to expose technical deficits but to introduce
conceptual frictions where the sector tends to stabilize its categories too
quickly. The Jade Emperor, Amaterasu, and Pamalak do not function as
pedagogical allegories, but as analytical devices aimed at denaturalizing
specific forms of order, visibility, and authority that are part of AI debates.
The Jade
Emperor challenges the assumption that the proliferation of instances,
mandates, and procedures automatically leads to greater accountability and, in
turn, produces more just algorithmic outcomes. In complex technical systems,
institutionalization does not eliminate the problem of responsibility but
redistributes it to diffuse it.
Amaterasu
and her mirror problematize the centrality of transparency and explainability.
“Light” does not in itself introduce justice, but may serve to re-legitimate an
order that had been weakened. The mirror returns the world to a functional
state, but not therefore to a more just or equitable one.
Pamalak
shifts the reading of anthropomorphism from a simple cognitive error or
literacy deficit to the realm of an objective intentionally pursued, in order
to blur the boundary between the human and the non-human through conceptual
confusion, so as to exploit its functional advantages of delegation, empathy,
trust, and authority, while simultaneously supporting the displacement of
responsibility for whatever may occur onto the model, the interface, or the
user.
Taken
together, the three myths leave us with a shared warning: many of the devices created to govern AI do not merely correct preexisting problems, but actively participate in instituting a particular technical and political order.
Castoriadis warned long ago that the social is not sustained solely by
functions or instrumental rationalities, but by instituted significations that
become self-evident to those who inhabit an order: “What we call ‘reality’ and
‘rationality’ are its works” (Castoriadis, 1987:9).
From this
perspective, the risk lies not only in the malfunctioning of systems but in the
progressive naturalization of a technical imaginary that presents certain
configurations of power as mere functional responses.
These myths
offer neither solutions nor the announcement of an inevitable collapse, but
they destabilize assumptions of technological objectivity and remind us that
every technical order is also an instituted, historical, and contingent order
of the distribution of roles and responsibilities—that is, a political order.
The most relevant question is not what limits AI should have, but what limits
we are willing to accept in our own practices to defend certain political
values entangled in the web of digital progress.
References
Ananny, M.
y Crawford, K. (2016). Seeing without knowing: Limitations of the transparency
ideal and its application to algorithmic accountability. New Media
& Society, 1-17. https://doi.org/10.1177/1461444816676645
Anesaki, M. (2015). Mitología japonesa: Leyendas, mitos y
folclore del Japón antiguo. Editorial Amazonia.
Benasayag, M. y Pennisi, A. (2024). La inteligencia
artificial no piensa (el cerebro tampoco), (5ª ed.). Prometeo.
Castoriadis,
C. (1987). The imaginary institution of society. Polity Press.
Coeckelbergh, M. (2023). La filosofía política de la
inteligencia artificial: Una introducción. Ediciones Cátedra.
Ellul, J.
(1964). The Technological Society. Vintage Books.
Foucault,
M. (2007). Security, territory, population: Lectures at the Collège de
France, 1977-1978, (M. Senellart, Ed.). Palgrave Macmillan.
Kroll,
J.A.; Huey, J.; Barocas, S.; Felten, E.W.; Reidenberg, J.R.; Robinson, D.G. y
Yu, H. (2017). Accountable Algorithms, University of Pennsylvania Law Review,
(3)165, 633-705. https://scholarship.law.upenn.edu/penn_law_review/vol165/iss3/3
Madrid Casado, C. M. (2024). Filosofía de la inteligencia
artificial. Pentalfa
Ediciones.
Parasuraman,
R. y Riley, V. (1997). Humans and automation: Use, misuse, disuse, abuse. Human
Factors, 39(2), 230-253. https://doi.org/10.1518/001872097778543886
Reeves, B.
y Nass, C. (1996). The media equation: How people treat computers,
television, and new media like real people and places. CSLI Publications;
Cambridge University Press.
Rudin, C.
(2019). Stop explaining black box machine learning models for high stakes
decisions and use interpretable models instead. Nature Machine Intelligence. https://doi.org/10.1038/s42256-019-0048-x
Storm, R. (2006).
Mythology of Asia
and the Far East: Myths and legends of China, Japan, Thailand, Malaysia, and
Indonesia.
Southwater.
Weizenbaum,
J. (1966). ELIZA - A computer program for the study of natural language
communication between man and machine. Communications of the ACM, 9, 36-45.
https://web.stanford.edu/class/cs124/p36-weizenabaum.pdf
Winner, L.
(1986). The whale and the reactor: A search for limits in an age of high
technology. University of Chicago Press.