Fleming and Michel’s hypothesis about the origins of consciousness not only faces multiple challenges but also overlooks phenomenal experience—the very phenomenon that needs explaining. Although we are in broad agreement with the authors’ overall perspective, we highlight challenges and suggest that consciousness evolved to attribute subjective value, enabling organisms to care about and act upon what they sense.

This comment explores resource-rational contractualism (RRC) as a framework for artificial intelligence (AI) alignment and reasoning. It assesses RRC’s capacity to capture the insights of philosophical contractualism and canvasses its extension to AI alignment. It suggests that a modified version of RRC, centered on impartial deliberation, can be used to identify principles for alignment and to guide efficient AI moral cognition.

This comment critiques the concept of “sensory horizon” for focussing narrowly on vision, overlooking aquatic animals’ rich multisensory integration. Aquatic vertebrates possess sophisticated auditory, mechanosensory, olfactory, and tactile systems that support essential behaviours. Evidence indicates that multisensory integration evolved in water, challenging visual-centric views. We thus urge replacing “sensory horizon” with “multisensory horizon” to more accurately reflect ecological and evolutionary realities.

Resource-rational contractualism claims that moral flexibility leads to mutually beneficial agreements through renegotiation when new situations arise. We challenge this view using COVID-19 evidence, where moral change leads to persistent disagreement rather than consensus. We demonstrate that three psychological factors – social identity, confirmation bias, and emotional responses – systematically prevent convergence, revealing limitations in the resource-rational contractualism model.

Levine et al.’s resource-rational contractualism (RRC) models moral judgment as a cost-sensitive approximation to ideal bargaining among stakeholders. We identify three strategies agents use to reduce moral costs by outsourcing or bypassing virtual bargaining: strategic delegation, diffusion of responsibility, and coalitional boundary-setting. We propose extending RRC with parameters for agency transfer, responsibility tracking, and stakeholder inclusion to better capture real-world moral decision-making.

Fleming and Michel propose that conscious vision evolved as a result of the water-to-land transition of animal life, enabling what they call “reality monitoring.” Here, we challenge their account by drawing on evolutionary arguments and the sensory capacities of aquatic life. Consciousness, we suggest, may have deeper evolutionary roots with significant roles for “slow” decision-making.

The target article maintains that consciousness emerged only when slow perceptual processing would be useful. But the formation of a conscious percept could be fast even if the vulcanization of that percept is slow. And fast formation may be too fast for a higher-order pointer to form.

The target article describes social cognitive mechanisms that enable people to make reasonable decisions in situations of interdependent choice where parties have divergent interests. We highlight the need for an integrative program of research on the social-cognitive mechanisms that people use to approximate the reasonableness standard of sound judgment via self-distancing and “inner dialogue” in a variety of decision-making contexts.

The authors make common assumptions that I reject: that consciousness has functions distinct from those of the brain processes on which it depends, that there is a moment at which consciousness happens, and that consciousness itself can be measured. This leads to Cartesian materialism and asking such unanswerable questions as when and why consciousness evolved. An alternative view is proposed.

Contractualism is valuable, but morality is not a means to agreements for mutual benefit. Beyond what we expect others to agree to, morality is – by almost any definition – about what we think others ought to agree to. Children, youth, and adults form judgments about social inequalities based on reasoning about rights and justice, whether they expect agreement or not.

In a vision-first story, conscious vision evolved before other kinds of conscious experience. This can be contrasted with an olfaction-first story, in which conscious vision co-opted integrative mechanisms that first evolved for modeling the causes of olfactory stimuli. An olfaction-first story makes good sense of the connection between consciousness and holistic integration across temporal windows in the ∼400ms range.

Levine et al. offer a compelling theory of moral decision-making, yet underemphasize the role of social perception. We argue that moral judgments function not only to evaluate actions but also to signal moral character. This signaling shapes social interactions and guides moral processing, highlighting the need to integrate reputational and relational dynamics into models of moral cognition.

What exactly is a contract? In this commentary, we discuss the need for a formal theory of how contracts are represented in the mind. Drawing inspiration from the representation of real-world contracts, we suggest that probabilistic programs are an appealing candidate representation.

A strict lower limit of 400 ms is suggested for the formation of conscious vision. This processing delay could result from the need to create a unified percept over visual features processed with different time constants, and even longer intentional visual processes. Given the flexibility of these processing times, we question the need for a strict lower limit for conscious vision.

Levine et al.’s resource-rational contractualism omits how people decide that a situation demands rule-, not case-based, guidance. I propose that moral cognition first tags contexts as rule-governed, then either uncovers an existing norm or forges a plausible one. Coordination stakes and the private-public divide likely trigger this tag. Distinguishing discovery from invention invites targeted experiments and refines the framework.

We challenge Fleming and Michel’s claim that conscious vision is too slow for real-time action. Correlative and functional evidence suggest that conscious perception can emerge around 200 ms post-stimulus and guide behavior. Due to their reliance on rare postdiction effects and assumptions about unconscious vision, they underestimate the speed of conscious vision and overestimate the functional relevance of unconscious vision.

The postdictive experimental paradigm is not superior to the direct detection paradigm, according to which the best strategy to identify the neural correlate of visual consciousness is to treat postdiction as a confounding factor. We could at most establish an equal claim to legitimacy of the two paradigms; hence, there is no puzzle about the function of a “slow” consciousness.

Bargaining requires agents to negotiate normative principles amid disagreement, incomplete information, and shifting social contexts, making uncertainty about moral rules and frameworks inherent to the decision process. In this commentary, I highlight the need for understanding how individuals resolve moral disagreements not just by applying fixed principles, but by navigating the uncertainty of which moral framework to adopt.

Fleming and Michel argue that consciousness lags up to 450 ms behind the world. I argue that the “aquatic sensory volume” for objects ecologically relevant for fish lies outside this window. If “reality monitoring” is an important function of consciousness, it likely evolved well before the transition to land in the vertebrate lineage.

We agree with Levine et al. that moral judgments track outcomes that would emerge from a balanced bargaining process. But we challenge the claim that moral cognition must therefore simulate bargaining itself. Like any evolved system that exploits deep regularities without representing them, moral cognition can produce ideal contractualist judgments without mentally representing bargaining scenarios.