Home > Positioning > Persons > Levin

Michael Levin (b. 1969)

Levin studies how bodies know what shape to be. His central finding is that resting membrane-voltage patterns across ordinary, non-neural tissue act as instructive signals for growth and form — a bioelectric layer of control above the genome, readable and rewritable. From that base his lab has produced some of the most striking results in modern biology: flatworms whose target anatomy is rewritten so that they regenerate two heads, and keep doing so when re-cut, with the genome untouched; a voltage “prepattern” of the future face visible on a frog embryo before any facial structure forms; and living constructs — xenobots from frog cells, anthrobots from adult human cells — that assemble, move, and in one case replicate, without any genetic modification. Around the experiments he has built a theoretical frame in which cognition is not a binary owned by brains but a continuous scale of agency, with tissues navigating anatomical space the way animals navigate physical space.

Life

Born in Moscow in 1969; the family emigrated in 1978 under the Soviet-Jewish visa program and settled in Lynn, Massachusetts — his father a programmer, his mother a concert pianist. Degrees in computer science and biology at Tufts, then a PhD at Harvard under Clifford Tabin, where his graduate work on the molecular basis of left-right asymmetry (1995) was later named a milestone in developmental biology. After a postdoc at Harvard Medical School and an independent lab at the Forsyth Institute, he returned to Tufts, where he is Vannevar Bush Distinguished Professor and directs the Allen Discovery Center, with an associate faculty position at Harvard’s Wyss Institute. More than four hundred publications span developmental biology, computer science, and philosophy of mind — the computer-science training visible throughout, in a biology done in the vocabulary of memory, encoding, and control.

The bioelectric programme

The bioelectric code. Cells throughout the body — not just neurons — maintain resting voltage differences across their membranes, and patterns of that voltage across tissue carry instructive information for anatomy: which organs form, where, and at what size. The lab reads these patterns with voltage-reporter dyes and rewrites them with ion-channel drugs and misexpression, in frog, zebrafish, fly and mammalian systems.

Pattern memory. The flagship demonstration is the two-headed planarian: briefly perturbing bioelectric signalling in a cut flatworm yields regeneration with a head at each end — and when the two-headed worm is cut again, with no further treatment, it regenerates two-headed again. The target morphology itself has been rewritten and is stably held in the tissue’s bioelectric state, while the genome remains unchanged — anatomical memory without neurons.

The electric face. In the frog embryo, voltage and pH reporters reveal a bioelectric prepattern on the ectoderm that matches the future arrangement of the face before any of its structures exist; disrupting the pattern disrupts craniofacial development. The pattern is there before the anatomy — a specification, not a shadow.

Xenobots and anthrobots. With Josh Bongard’s AI lab, the group built xenobots (2020): millimetre-scale motile constructs of frog skin and heart cells whose body plans were evolved by algorithm and assembled by hand — followed in 2021 by kinematic self-replication, xenobots sweeping loose cells into piles that mature into new xenobots. Anthrobots (2023) made the point without any design step: adult human tracheal cells, unedited, self-assemble into motile ciliated bots that in vitro encouraged healing across scratched sheets of neurons. The constructs show ordinary cells doing extraordinary things given a new context — capabilities latent in the material, not engineered into it.

The frame. “Cognition all the way down” (with Daniel Dennett, 2020) argues that cells have goals of a modest, cell-sized kind, and that larger minds are collective intelligences with larger goal-spaces. TAME — the Technological Approach to Mind Everywhere (2022) — makes the frame operational: agency as a continuous scale, measured by what Levin calls the cognitive light cone, the size of the largest goal a system can represent and pursue; morphogenesis becomes collective problem-solving, tissues correcting their way toward an anatomical target through a space of possible forms.

The live dispute. The reception splits cleanly. The bioelectricity results are mainstream science — published in the field’s central journals, well funded, independently extended. The pushback targets the cognitive vocabulary: critics such as Massimo Pigliucci argue that goal-and-cognition talk about cells is metaphor doing no explanatory work, and later, more speculative extensions of the frame have drawn their own criticism. The experiments are not in dispute; what they license is.

Where Levin stops

His answer to “is it really cognition?” is to decline the question’s form. On the TAME framing there is no metaphysical yes/no to settle — only positions on a continuous scale of agency, and the operational question of which level of description yields the best prediction and control of the system at hand. That is a considered boundary, not an evasion, and he polices it himself: TAME is offered as a framework for generating experiments, not a doctrine; he explicitly disclaims panpsychism and makes no claims about phenomenal consciousness in cells. What the programme deliberately does not supply is what its critics keep asking for — a principled line where agency-talk must stop. On Levin’s own terms no such line exists to be drawn; whether that dissolves the demarcation question or merely postpones it is exactly where his frame and its critics part company.

Key works

See also: Lyon · Dennett · Maturana · Basal cognition