Claim № i.01  · Lens · Case file

“Natural processes, chiefly mutation and natural selection, can generate the functional, specified information found in living things.”

Disciplines: Information theory · Computer science
Status: reviewed 2026-11-15
UnderExaminationTwo claims · two grades

§1 The claim, as taught

Some mutations add information to a genome; some subtract it. … By any reasonable definition, increases in information have been observed to evolve.Isaak M (ed.). *Index to Creationist Claims*, TalkOrigins Archive, claim CB102, “Mutations adding information.”

§2 The strongest evidence offered

Here the argument is as much about concepts as about data. We set out what each side means by information, and what has been shown.

Inferred

Three meanings

Claude Shannon's measure counts how improbable a sequence is, not what it means: a random string scores highest. Functional information, as defined by Jack Szostak and Robert Hazen, measures how rare the sequences are that perform a given function. Both differ from the information of a code, in which symbols stand for something else.
Shannon, Bell System Technical Journal 27 (1948); Szostak, Nature 423 (2003); Hazen et al., PNAS 104 Suppl. 1 (2007).
Observed

Selection increases information

In Thomas Schneider's simulation ev, binding sites evolved from random sequences under selection, and their Shannon information rose to match what the recognising protein needed. In the lab, new binding and regulatory functions have arisen under selection.
Schneider, Nucleic Acids Research 28 (2000)
Observed

Complex functions in silico

In the Avida system, digital organisms evolved a complex logic function, EQU, through many intermediate steps, but only when simpler functions were also rewarded.
Lenski, Ofria, Pennock & Adami, Nature 423 (2003)
Inferred

Searches need help

William Dembski and Robert Marks argued, building on the no-free-lunch theorems of computer science, that a search performs better than chance only if information about the target is supplied by its design. Critics reply that the real environment, not a programmer, supplies that information.
Wolpert & Macready, IEEE Trans. Evol. Comput. 1 (1997); Dembski & Marks, IEEE Trans. SMC-A 39 (2009); Elsberry & Shallit, Synthese 178 (2011).
Inferred

The design inference

Stephen Meyer argues that the only cause known from experience to produce large amounts of specified, functional information, such as software, writing or engineered systems, is intelligence, so that intelligence is the best explanation for the information in DNA.
Meyer, Signature in the Cell (2009); Dembski, The Design Inference (1998).

§3 The grade: two claims hiding in one

Small, measurable gains and the origin of life's functional information are different claims. Graded separately: Hover or tap a rung to see why it got its score. How the grading works →

Mutation and selection can increase information in a sequence, in measurable ways

BStrongly supported
Observation
///
Observed in simulations and in lab selection for binding and regulatory functions.
Hypothesis
///
Precise definitions and measures.
Prediction
//
Predicted gains are found where selection rewards the function.
Experiment
//
Controlled simulations and lab experiments.
Replication
///
Replicated.
Falsifiability
//
Testable, within the systems studied.
Method score 15/18

Unguided processes produced the large amounts of functional information in genes and proteins

DExtrapolation
Observation
/
Gains observed are small; the origin of new protein folds and codes is not observed.
Hypothesis
//
A clear hypothesis.
Prediction
/
Estimates of functional rarity (Pillar 4) imply searches too large for the resources available.
Experiment
/
Simulations that succeed provide graded paths to their targets.
Replication
/
No replicated demonstration at the scale required.
Falsifiability
/
Failures of particular simulations are not taken to count against the general claim.
Method score 7/18
ADemonstratedBStrongly supportedCInferenceDExtrapolationEAssumption

§4 What it would take

To show that unguided processes can generate life's functional information, the following would be needed:

  • A new fold, unaidedObserve a new protein fold, or a new multi-part system, arising by mutation and selection without a designed fitness landscape.
  • A simulation without a mapRun an evolutionary simulation whose fitness function does not encode the target or the path to it, and show that it reaches comparable complexity.
  • Measure, then compareMeasure the functional information of real biological systems and compare it with the amounts mutation and selection have been seen to produce.

§5 Another reading of the same data

The debate about information often goes wrong at the first step, because the two sides mean different things by the word. If information means Shannon information, then mutation creates it trivially: any random change makes a sequence less predictable. If it means the ability to bind one particular molecule, selection has been seen to produce it. Neither of those is what design advocates have in mind.

What they mean is information of the kind found in a protein-coding gene: a sequence that is both highly improbable and specified by a function, so that very few alternatives would work. Pillar 4 found that, on some careful estimates, working protein folds are extraordinarily rare among possible sequences. The simulations that succeed in evolving complex functions, such as Avida, succeed because the programmer laid out graded rewards along the way. That is a fair test of a search with a map; it is not yet a test of a search without one.

Read through a design lens, DNA is the most striking case of a code we know: a system in which sequences of symbols specify, through a translation machine, the structure of something else. In every case where we know how a code and its information came to be, a mind was involved. That is not proof, and it would be overturned by a demonstration of a code or a fold arising without one. It is an inference from what we know, not from what we do not, and it is where every pillar's open question points.

§6 Reasoning check

The grade weighs the evidence. This check looks at the reasoning: the step from the evidence to the conclusion. The same tests are applied to both readings. How the reasoning check works →

In the claim as taught
  • Clear case

    Gains in Shannon information or in a single binding function are presented as the same kind of thing as the information that specifies protein folds and the genetic code.

  • Possible

    From small information gains, in simulations or in binding sites, to the origin of whole genes, folds and systems.

  • Possible
    In the evidence: Complex functions in silico

    Avida's reward schedule was designed so that simpler logic functions led stepwise to EQU. It shows that a search succeeds when the path is laid out. Whether biology supplies such paths is the question at issue. The authors also showed that EQU did not evolve when the stepping stones were not rewarded.

In the other reading
  • Possible

    Critics say the design inference argues from what evolution has not done. Its defenders reply that it argues from what intelligence is known to do. Both premises need support: the first from failed searches, the second from our uniform experience of codes and machines.

§7 Questions to ponder

  • A random string has more Shannon information than a sentence. What kind of information do we mean when we talk about DNA?
  • If a simulation succeeds because its designer rewarded the right stepping stones, what has it shown?
  • Is “intelligence is the only known cause of specified information” an argument from ignorance, or from knowledge?

§8 Sources

Supporting the mainstream claim
  1. Shannon CE. “A mathematical theory of communication.” Bell System Technical Journal 27:379–423 (1948).A
  2. Schneider TD. “Evolution of biological information.” Nucleic Acids Research 28:2794–2799 (2000).B
  3. Lenski RE, Ofria C, Pennock RT, Adami C. “The evolutionary origin of complex features.” Nature 423:139–144 (2003).B
  4. Szostak JW. “Functional information: molecular messages.” Nature 423:689 (2003).C
  5. Hazen RM, Griffin PL, Carothers JM, Szostak JW. “Functional information and the emergence of biocomplexity.” PNAS 104 Suppl. 1:8574–8581 (2007).B
  6. Wolpert DH, Macready WG. “No free lunch theorems for optimization.” IEEE Transactions on Evolutionary Computation 1:67–82 (1997).B
  7. Elsberry W, Shallit J. “Information theory, evolutionary computation, and Dembski's ‘complex specified information’.” Synthese 178:237–270 (2011).C
Critical & alternative perspectives
  1. Dembski WA. The Design Inference: Eliminating Chance through Small Probabilities. Cambridge University Press (1998).C
  2. Dembski WA, Marks RJ II. “Conservation of information in search: measuring the cost of success.” IEEE Transactions on Systems, Man, and Cybernetics A 39:1051–1061 (2009).C
  3. Meyer SC. Signature in the Cell: DNA and the Evidence for Intelligent Design. HarperOne (2009).C

Source grades are provisional. In production each links to a full published evaluation that readers can challenge.

§9 Your notes