Phase Space of Mind: Why Israel Can’t Afford to Ignore Consciousness Questions
When an autonomous loitering munition selects and engages a target without human intervention, we assume we know what it is not: it is not aware. It does not experience the moment of impact. It does not, in any meaningful sense, choose. But the confidence behind that assumption rests on a question that neither Israeli defence procurement nor anyone else has rigorously answered: what is consciousness, and at what point, if any, does a sufficiently complex artificial system acquire it?
Israel’s economic model and security architecture depend more heavily on artificial intelligence than those of almost any other democracy. Rafael’s AI-driven weapons platforms, Mobileye’s perceptual networks, the intelligence systems fed by the Unit 8200 pipeline, and the proliferating ecosystem of Israeli AI ventures are growing in complexity at a pace that outstrips the frameworks meant to govern them. The consciousness question is not, for Israel, a seminar topic. It is a planning contingency.
Three competing models define the debate. The first treats consciousness as binary: you have it or you do not. This underwrites most legal codes, where personhood is conferred categorically, never by degree. The second treats it as occupying qualitatively distinct states separated by sharp transitions, much as water becomes ice at a critical temperature. Giulio Tononi’s Integrated Information Theory, which quantifies consciousness through a metric called phi, supports this view: the brain under anaesthesia does not drift gently toward unconsciousness but undergoes what Tononi’s own research describes as a breakdown of information integration. The third model is the pure continuum: consciousness as a fundamental property of any system that integrates information, present in vanishing degree even in a photodiode.
Each model carries different implications, but they converge on one point that matters for Israeli policy: if consciousness is not exclusively biological, then substrate independence means it could, in principle, emerge in silicon. The question is not whether today’s systems are conscious. They almost certainly are not. The question is whether the trajectory of complexity leads toward a threshold, and whether we will recognise that threshold when it arrives.
Consider the operational reality. IAI’s Harpy loitering munition already exercises fully autonomous target selection against radar emitters, without human intervention after launch. The Harop maintains a man in the loop, but the trajectory across successive generations of autonomous systems is unmistakably toward greater machine autonomy in target discrimination and engagement. The next generation of such platforms will incorporate neural network architectures whose internal representations their designers cannot fully inspect. If consciousness is binary and biological, this presents no ethical dilemma. If it is not, then a system capable of autonomous perceptual integration, target discrimination, and adaptive decision making under uncertainty begins to approach the functional criteria that some theorists consider markers of awareness. We need not believe these systems are conscious today to recognise that the absence of a rigorous test for machine consciousness leaves a dangerous gap in governance.
This gap matters more for Israel than for nations whose AI deployments are less operationally consequential. A European chatbot that might or might not possess rudimentary awareness raises philosophical questions. An autonomous weapons platform that might or might not possess rudimentary awareness raises questions of international humanitarian law. Although Israel is not party to Additional Protocol I to the Geneva Conventions, most of its weapons export customers are, and the Protocol’s Article 36 requirement that states review new weapons for legal compliance increasingly encompasses questions of autonomy. No existing weapons review framework, in Israel or elsewhere, includes a consciousness assessment.
Israel is well placed to close this gap, and the mechanism is not mysterious. The Ministry of Defence should establish a standing interdisciplinary commission, drawing on the Hebrew University’s Edmond and Lily Safra Center for Brain Sciences, the Weizmann Institute’s neuroscience programme, and the Technion’s AI research, and tasked with developing substrate-independent consciousness assessment protocols. Such a commission should be embedded within the weapons procurement review process, so that every autonomous system above a defined threshold of decisional complexity undergoes a formal consciousness evaluation before deployment. This is not a philosophical indulgence. It is the same logic that already requires environmental and electromagnetic compatibility assessments: identify the risk before it materialises, not after.
The honest admission is that we do not yet possess a validated, observer-independent measure of consciousness applicable across substrates. We cannot compute Tononi’s phi for systems of realistic complexity. But the absence of a perfected instrument has never stopped Israel from building interim frameworks for emerging threats. The country’s cyber directorate was established years before the threat landscape it now manages had fully taken shape. Consciousness governance demands the same anticipatory posture. The possibility that artificial awareness could emerge in systems designed to make life and death decisions falls squarely within the doctrine of preparing for threats whose probability is uncertain but whose consequences are severe.
Getting the measurement right before getting the ethics wrong is not a luxury. For a democracy whose security depends so fundamentally on the technology in question, it is a strategic obligation.