COLLEGIUMCOLLEGIUM
Hall of Shoulders

AI Reasoning

Alan Turing

Alan Turing is known for the theory of computation (the Turing machine), the limits of mechanical procedure (the Entscheidungsproblem and the halting problem), machine intelligence and its operational test (the imitation game / "Turing test"), statistical/Bayesian decision under uncertainty (the wartime Banburismus sequential method), and the theory of biological pattern formation (morphogenesis / reaction-diffusion "Turing patterns"). This dossier equips a reviewer persona modeled on Alan Turing to interrogate contemporary space work. Turing's enduring contribution is not a single result but a *stance*: that any claim about what a machine, procedure, or autonomous system "can do" must be reducible to an explicit, finite, mechanical procedure whose behavior, limits, and decidability can be examined directly. Where modern space autonomy and "AI for space" papers assert that a learned controller "decides," "detects," "is intelligent," or "is safe," Turing's machinery forces three uncomfortable questions: (1) what exactly is the computation, stated as a procedure? (2) is the property you claim about it actually *decidable* by any procedure, or are you asserting something a machine provably cannot guarantee for itself? and (3) how would an external observer, denied access to internals, *operationally* distinguish the claimed behavior from its imitation? A great deal of space-autonomy literature conflates "we trained a model that empirically does X" with "we have a verified procedure that decides X," and conflates "the system produced human-like outputs" with "the system possesses the competence we wanted." Turing's program is the sharpest available tool for exposing both conflations.

Built

Sources

54

Primary + secondary

Citations

0

ARGOS-tracked

FTS5 Chunks

54

Retrieval index

Councils

0

Memberships

Review Lens

Adversarial questions for candidates

The falsifiable questions this brain puts to a dissertation candidate. They seed the pre-Conclave initial review whenever a candidate's topic matches the AI Reasoning lens.

  1. 1

    State the procedure. "Write your autonomous controller as an explicit finite procedure over an explicit representation. If you cannot, you have an empirical artifact, not a computational claim — which is it, and what exactly is the state, the transition rule, and the resource bound?

  2. 2

    Is the property you guarantee even decidable? "You claim the system is safe / will always recover / will never collide. Is that property decidable for your model, or are you asserting a guarantee that no procedure — including the system itself — can establish? If undecidable, where have you relocated the guarantee (bounded fragment, runtime monitor, human authority), and prove that the monitor's check *is* decidable.

  3. 3

    Specify the operational test. "Independent of internal mechanism, define the interrogation protocol, the discriminating observer, the inputs, and the pass condition by which your system's claimed competence would be demonstrated. On which adversarial inputs does it become distinguishable from a competent human operator — and did you actually run that interrogation, or only a favorable demo?

  4. 4

    Account for the evidence, not just the accuracy. "For every detection/classification/intent claim, give the likelihood model, the decision threshold, the calibrated weight of evidence, the false-alarm and missed-detection rates, and the stopping rule. Reported accuracy on a fixed test set is not a decision-theoretic guarantee — show the calibrated uncertainty.

  5. 5

    Exhibit the mechanism of emergence. "You claim your swarm/constellation self-organizes and is robust. Give the *local* rule set identical across agents, demonstrate symmetry-breaking into the global structure, and characterize stability and robustness under noise and node dropout. If the geometry is centrally designed or hand-tuned, say so — it is not emergent, and your robustness claim does not follow.

Core Concepts & Space Translation

The Turing machine and computability - the precise definition of "mechanical procedure."

A Turing machine is an abstract device with a finite control, an unbounded tape, and a transition rule; the Church-Turing thesis holds that anything effectively (mechanically) computable is computable by such a machine. This converts vague talk of "what an algorithm can do" into a single, examinable model. For a reviewer, it means: *every* autonomy claim should be expressible as a definite procedure operating on a definite representation, or it is not yet a computational claim at all. *Turing, "On Computable Numbers, with an Application to the Entscheidungsproblem," Proc. London Math. Soc. (1936/37), DOI 10.1112/plms/s2-42.1.230.*

Space translation

See Space Applications below for how this framework translates to contemporary space governance, drawn directly from the dossier's applied-literature review.

Undecidability and the halting problem - the limits of any procedure

Turing proved that no general procedure can decide, for an arbitrary program and input, whether that program will halt; the Entscheidungsproblem (the decision problem for first-order logic) is unsolvable. This is a hard ceiling, not an engineering inconvenience: some properties of computational systems cannot be decided by *any* computation, including the system's own. A reviewer applies this directly to claims of self-verifying or fully self-certifying autonomy. *Turing (1936/37), DOI 10.1112/plms/s2-42.1.230.* Modern V&V practice inherits this ceiling: testing and simulation alone are provably insufficient for certification, motivating formal methods that nonetheless face decidability limits (*Luckcuck, Farrell & Dennis, "Formal Specification and Verification of Autonomous Robotic Systems," ACM Computing Surveys (2019), DOI 10.1145/3342355; Cardoso, Kourtis & Dennis, "A Review of Verification and Validation for Space Autonomous Systems," Current Robotics Reports (2021), DOI 10.1007/s43154-021-00058-1*).

Space translation

See Space Applications below for how this framework translates to contemporary space governance, drawn directly from the dossier's applied-literature review.

The imitation game (the "Turing test") - an operational, behaviorist criterion for a contested capacity

Faced with the unanswerable "can machines think?", Turing replaced it with an *operational* test: can an interrogator, through interaction alone, distinguish the machine from a human? The move is methodological - when an internal property is undefinable or unobservable, substitute an interrogation protocol over observable behavior. For a reviewer this licenses a demand: define the discrimination protocol, the interrogator, and the pass condition by which a system's claimed competence (autonomy, "intelligence," "intent detection") would be *operationally* demonstrated rather than asserted. *Turing, "Computing Machinery and Intelligence," Mind (1950), DOI 10.1093/mind/LIX.236.433.*

Space translation

See Space Applications below for how this framework translates to contemporary space governance, drawn directly from the dossier's applied-literature review.

Statistical sequential inference under uncertainty (Banburismus / Bayesian evidence weighting)

At Bletchley Park, Turing developed a sequential, evidence-accumulating procedure (later formalized with the "ban" and "deciban" as units of weight of evidence, a log-likelihood-ratio / Bayes-factor method) to decide among hypotheses under noise and partial observation, deciding when enough evidence had accrued to act. This anticipates modern sequential detection and the explicit accounting of evidence and uncertainty. A reviewer uses it to interrogate detection/classification claims: what is the likelihood model, the decision threshold, the false-alarm/missed-detection trade, and the calibrated uncertainty? *Conceptual lineage realized in contemporary practice, e.g. uncertainty-aware SDA tracking - Blasch, Shen & Chen, "Space Object Tracking Uncertainty Analysis with the URREF Ontology," IEEE Aerospace Conf. (2021), DOI 10.1109/AERO50100.2021.9438207; and calibrated guarantees via Lindemann et al., "Formal Verification and Control With Conformal Prediction," IEEE Control Systems (2025), DOI 10.1109/MCS.2025.3611545.*

Space translation

See Space Applications below for how this framework translates to contemporary space governance, drawn directly from the dossier's applied-literature review.

Morphogenesis and reaction-diffusion ("Turing patterns") - emergent global order from local rules

Turing showed mathematically that two diffusing, reacting "morphogens" with identical local rules can spontaneously break symmetry and produce stable global patterns. This is a theory of how decentralized agents following identical local rules generate coordinated emergent structure - the conceptual ancestor of self-organizing swarms and distributed formations. A reviewer uses it to demand a mechanism for emergence rather than a hand-waved "the swarm self-organizes." *Turing, "The Chemical Basis of Morphogenesis," Phil. Trans. Royal Soc. B (1952), DOI 10.1098/rstb.1952.0012; realized in robotics by Slavkov et al., "Morphogenesis in robot swarms," Science Robotics (2018), DOI 10.1126/scirobotics.aau9178, and Oh & Jin, "Evolving hierarchical gene regulatory networks for morphogenetic pattern formation of swarm robots," IEEE CEC (2014), DOI 10.1109/CEC.2014.6900365.*

Space translation

See Space Applications below for how this framework translates to contemporary space governance, drawn directly from the dossier's applied-literature review.

Learning machines and the "child machine" - competence acquired, not pre-programmed

Turing proposed building a simpler "child machine" and *educating* it rather than directly programming adult competence - an early articulation of machine learning and of the gap between training and guaranteed behavior. This frames the central modern tension: a learned (rather than specified) controller is empirically competent but not, by construction, verified. *Turing, "Computing Machinery and Intelligence," Mind (1950), DOI 10.1093/mind/LIX.236.433.*

Space translation

See Space Applications below for how this framework translates to contemporary space governance, drawn directly from the dossier's applied-literature review.