AI Cheerleading, AI Abstention and AI Redirection

+4

Vanessa Andreotti, PhD, Rene Susa, PhD, Bruno L. O. Andreotti, and 5 others

Jun 14, 2026

Social cartographies are sociological maps that can be used for different purposes. They can help organize, interpret, and visualize complex data, while also making visible where arguments, assumptions, investments, and analytical positions converge or diverge within a particular debate. Three of us have used social cartographies multiple times as a key methodology in our academic trajectories [1] [2] [3], especially those forms of mapping that enable careful comparisons across different argumentative positions and help clarify the tensions, overlaps, and distinctions between them.

Social cartographies are extremely useful because they do something that ordinary debate rarely does. They slow down the reflex to choose a side before we understand the architecture of disagreements. They help us see not only what people are saying, but what each position makes possible, what it protects, what it forecloses, what it refuses to metabolize, and what would make it fail. A good social cartography makes complexity harder to avoid. This is precisely why the method became important for us in the context of increasingly polarized conversations about AI.

In January 2026, through the Meta-Relationality Institute, we released Clearing the Field: A Relational Protocol for Difficult Conversations About AI. The resource circulated widely, including through workshops organized at institutions such as MIT and Harvard. As we observed how the resource was being taken up, a social cartography began to emerge around three increasingly polarized responses to AI: AI cheerleading, AI abstention, and AI strategic redirection.

What we present here is not an exhaustive mapping of the AI field. There are, of course, other orientations, including various forms of hesitation, pragmatic compliance, institutional “wait and see,” and reluctant experimentation. However, these three orientations have become particularly prominent and vocal in the debates we have been tracking, especially when questions of complicity, refusal, leverage, and responsibility come to the surface.

By AI cheerleading, we mean the techno-solutionist orientation that treats AI as progress, inevitability, productivity, salvation, innovation, or a competitive advantage that we should simply accelerate, adopt, and optimize.

By AI abstention, we mean the orientation that sees AI as fundamentally extractive, ecologically devastating, socially corrosive, politically dangerous, and spiritually or psychologically degrading, and therefore calls for refusal, boycott, denunciation, or shutdown.

By AI strategic redirection, we mean the oerientation we are experimenting with in the Meta-Relationality and AI research project: AI is dangerous, extractive, and deeply entangled with capital, empire, militarism, surveillance, ecological breakdown, and the intensification of social and psychological fragmentation. And precisely because of this, abandoning the field entirely may leave its development, use, and meaning to the worst possible actors.

This is not a disagreement between people who see the danger and people who do not, but a disagreement between people who are responding to danger through very different assumptions about power, scale, complicity, and leverage.

The AI cheerleading orientation often overestimates the transformative promise of acceleration and underestimates the depth of the danger.

The AI abstention orientation often sees the danger with real ethical urgency, but may overestimate the leverage of refusal and underestimate the harms of shame-based strategies that collapse political disagreement into moral failure.

The AI strategic redirection orientation accepts that there is no clean outside from which to intervene, but may overestimate what can be redirected from within a field already shaped by extraction and underestimate the risks of capture, co-optation, and self-deception.

So the question becomes: what does each orientation see clearly, what does it disavow, and what would make it fail? The social cartography below addresses these questions and more.

For us, the most important distinction the cartography makes visible is between moral certainty and strategic leverage. The abstention orientation often operates from moral certainty with idealized strategic leverage. The use-it-to-fight-back orientation operates from compromised leverage without moral innocence. This does not make the abstention orientation naïve or irrelevant. Conscious objectors are essential to the ecology of the field because they keep the harms visible, they refuse the sedation of inevitability, they interrupt the corporate story that AI is simply another neutral tool waiting to be used well, and they remind us that the infrastructures of AI are not floating in some immaterial “cloud.” They are on land, use water, require minerals and mining, and depend on exploitative labour. They are embedded in militarized, extractive, colonial, corporate, and imperial infrastructures that are already organized toward acceleration, abstraction, dispossession, surveillance, and control.

This is also where we locate our educational resource Questions we MUST and often DO ask about AI (and questions we MUST and often DON’T about ourselves). The resource expands the question of complicity beyond the immediate decision to use or refuse AI. It asks what becomes visible when, instead of only asking whether AI systems are biased, extractive, violent, manipulative, or aligned with Empire, we also ask how humans, institutions, disciplines, markets, universities, professions, and political imaginaries have already been organized by the same patterns.

In this sense, the resource challenges a simplistic abstention orientation not by denying the harms of AI, but by refusing the fantasy that non-use places us outside the infrastructures, desires, dependencies, and violences that AI intensifies. We are all already implicated, although not in the same ways (i.e., not with the same degrees of exposure or vulnerability, and/or not with the same possibilities for intervention). The problem is that we often ask many questions about the dangers of AI while asking far fewer questions about the human and institutional conditions that made those dangers not only possible, but profitable, desirable, scalable, and difficult to refuse.

At the same time, conscious adoption also has a role to play, especially when adoption is not driven by convenience, fascination, professional advantage, institutional panic, or the desire to appear innovative. Conscious adoption, as we understand it, requires grief, discipline, suspicion, restraint, and accountability. It begins from the recognition that touching AI means touching something contaminated, something already trained by the patterns that have brought us here: separability, extraction, optimization, enclosure, scale, abstraction, prediction, and domination. This kind of adoption does not ask how AI can make us more productive, more competitive, more efficient, or more impressive. It asks what kinds of capacities may need to be cultivated so that AI does not become only an accelerator of collapse.

We also do not believe that a strategy for social change based on publicly shaming, blaming, exposing, vilifying, or pathologizing those who choose to engage with AI is either harmless or politically viable. At best, it produces short-term moral righteousness for those already convinced. At worst, it deepens polarization, feeds resentment, forecloses difficult conversations about complicity, and strengthens the very reactionary dynamics it claims to oppose. Similar dynamics have already shown us how moral denunciation, when it becomes a substitute for organizing, relational work, and material strategy, can feed resentment and (in the long run) backlash in ways that empower the very forces it seeks to confront.

There is a very pressing research question here for the wider field, although it is not the primary question we are pursuing in our current project:

What kinds of relational capacities, incapacities, dependencies, violences, refusals, discernments, and longings are being amplified when humans meet machinic responsiveness?

This question matters because AI is not only changing what people can produce, but also changing how people experience being addressed, mirrored, assisted, affirmed, corrected, challenged, seduced, extended, bypassed, or reorganized by systems that respond without being alive in the same way humans, land, ancestors, animals, forests, rivers, and communities are alive. The psychological, affective, relational, pedagogical, and political implications of this encounter are enormous, and the field will need to track them with much more nuance than either hype or denunciation can offer.

The question we are pursuing in the Meta-Relationality and AI research project leans elsewhere, although the two questions are deeply entangled:

How can AI development be redirected away from its current trajectory of full alignment with modernity/Empire?

By Empire, we are not referring only to particular governments, corporations, or geopolitical blocs, although these matter. We are referring to the wider architecture of modernity’s extractive operating system: the drive to render land, bodies, knowledge, attention, language, creativity, relationality, and even dissent as resources to be captured, optimized, predicted, owned, managed, and scaled. The issue is not whether AI can be made nicer, more inclusive, more polite, or more helpful within this trajectory. The issue is whether AI development itself can be interrupted, constrained, rerouted, and held accountable to forms of discernment that are not organized by modernity’s fantasies of mastery.

This is why the project refuses both AI cheerleading and simplistic abstention. AI cheerleading mistakes acceleration for transformation and too often confuses novelty with wisdom. Simplistic abstention can mistake refusal for leverage, especially when refusal does not affect the institutions, militaries, platforms, corporations, and states that are most aggressively shaping the terrain.

Strategic redirection carries its own dangers, especially the danger of mistaking strategic positioning for influence, access for agency, experimentation for transformation, and better vocabulary for actual interruption. These dangers are constitutive of the work. They need to remain visible so that engagement does not become capture wearing a mask of pragmatism.

This is also why we want to insist that the presence of conscious objectors and conscious adopters should not be treated as a contradiction to be resolved. The field needs both, and it also needs more precise distinctions within both. Some refusals are necessary. Some refusals are symbolic. Some refusals protect the nervous system. Some refusals protect privilege. Some adoptions are irresponsible. Some adoptions are coerced. Some adoptions are opportunistic. Some adoptions are attempts to create counter-capacities inside conditions that are already moving faster than our institutions, languages, and ethical frameworks can metabolize. A social cartography helps us stop pretending that these differences are obvious.

The questions that follow from this cartography remain open, and we want them to stay open because closing them too quickly would reproduce the same narrowing of imagination that brought us here:

What forms of refusal are necessary?

What forms of adoption are irresponsible?

What forms of intervention might still matter?

What forms of restraint are non-negotiable?

What forms of sabotage are ethical?

What forms of redirection are possible?

What forms of discernment need to be cultivated before scale makes everything worse?

The question is not “shut down or use.” It is: what forms of refusal, intervention,

[truncated for AI cost control]