Is the universe fine-tuned for life?

The short version

1. Fine-tuning is a measured sensitivity, not a mystery.

   A constant counts as “fine-tuned” only if physics gives both a life-permitting window and a natural range to compare it with.

   The rules ↓
2. The best case is the energy of empty space.

   In standard physics it drops by about 10⁵⁵ times its present value when the electroweak force separates. Galaxies and stars need what is left within roughly −100 to +5,000 times today's value, a deliberately generous window.

   The evidence ↓
3. Against blind chance in one universe, the evidence is overwhelming.

   If nothing links the vacuum's starting value to particle physics, a life-permitting result has a probability of at most about 1 in 10⁵¹, whatever prior you use, as long as it doesn't already aim at the answer.

   The calculation ↓
4. Against naturalism as a whole, it is a judgment call.

   A cancelling law, a multiverse with selection or a mind would each make the result expected. Which is likelier is a question of priors this evidence cannot settle.

   The rivals ↓
5. Suggestive evidence for a life-intending Mind.

   A likelihood ratio of about 10 (range 0.5 to 10³), down from our earlier 50. It points to a Mind in general, not to the God of the Bible in particular.

   What it shows ↓

The question

The evidence, E, is a fact about this universe: its laws, constants and starting conditions fall in a region that allows long-lived stars, a rich chemistry and so embodied, evolved observers, and that region is small compared with the range our best physical theories treat as natural. The question is which hypothesis makes E more expected.

Naturalism in the broad sense is a weighted mix of these. That turns out to be the crux: the argument is overwhelming against N0 and much weaker against the mix, because N1 and N2 make the observation just as expected as design does.

We test a likelihood ratio: how much more probable E is if D is true than if naturalism is true. It is not a “probability of God”. That would also need the prior odds, which belong to the unified argument.

Rules of evidence

Fixed before scoring anything:

1. Both numbers or nothing. A parameter counts as fine-tuned only if a published physics source gives a life-permitting window and a natural range from a stated theory or scale. A “percent of what?” with no answer is excluded.
2. Life, not Earth. “Life-permitting” means robust criteria: stars that last billions of years, stable atoms with more than one element, carbon and oxygen made in stars, cosmic structure. Day length and year length don't count.¹
3. Vary several dials at once. Each case records whether its window survives when other constants change too. Where a published study opens another habitable island by re-tuning other constants, the case is marked contested.
4. Always state the measure. Every probability comes with its prior (flat or logarithmic) and its cut-off. Where that choice moves the answer by orders of magnitude, we sweep it.
5. No multiplying across parameters unless independence is argued. The constants are linked (particles to nuclei to stars to chemistry), and one landscape would explain them all. The headline therefore rests on one parameter, the cosmological constant, because its measure is the best motivated.
6. Show the null space. The evidence must include the dials that turn out not to be finely tuned, and the evidence pointing the other way.
7. Planetary conditions are scored per universe, using measured occurrence rates, not per planet.
8. Penrose's initial-entropy number is reported but not scored, because the probability measure over initial states is itself disputed.
9. Every number carries its scope. A probability “for one universe by chance” is never presented as a probability “under naturalism”, and the same fact restated in bits is never counted as a second piece of evidence.

The evidence

The best case: the energy of empty space

The cosmological constant, written ρ_Λ, is the energy density of empty space (the “dark energy” that drives the accelerating expansion). Quantum field theory says empty space should carry energy from many sources: zero-point energies of every field, phase transitions such as the QCD and electroweak ones, and a bare term. They come with both signs, and they are about 10¹²⁰ times the observed total.¹

That sum matters for life. Too large and positive, and expansion accelerates before matter can clump, so there are few or no galaxies or stars. Too negative, and the universe recollapses before life has time. Simulations of galaxy formation with the EAGLE code found that raising Λ even tenfold has only a small effect, because star formation peaks before acceleration sets in. At 100 times the observed value, about 0.5% of the ordinary matter still ends up in stars, against about 4% in a universe with no Λ at all: an eightfold drop, not a stop.² Weinberg's original bound, as reported in reviews, sits around 225 times the observed value.³ So the window is roughly −10 to +100 times the observed value, and generous estimates stretch it to a few hundred. To be generous to chance we use −100 to +5,000. Raise the primordial ripples (Q) at the same time and it widens to about 10⁶.⁴

What the window has to cancel. In standard physics (general relativity with the Standard Model), the energy of the vacuum drops when the electroweak symmetry breaks, by about the Higgs mass squared times the Higgs field value squared, over 8. With measured values that is about 1.2×10⁸ GeV⁴, or 10^54.6 times today's vacuum energy. Martin's review gives the same: “ρ^EW_vac ≃ −1.2 × 10⁸ GeV⁴ ≃ −10⁵⁵ ρ_cri”.⁵ So whatever the vacuum energy was before the transition, it had to cancel that drop to about one part in 10⁵¹. Nor can the cancellation be built into both states at once. As Polchinski puts it, “It cannot vanish in both because the electron mass is zero in the symmetric state and not in ours”.⁶

Even raising Q along with Λ, a window of 10⁶ is only 10^−48.6 of the drop it must cancel. No re-tuning of other dials closes that gap. Only a mechanism (a landscape, a cancelling law) could, or a theory of gravity in which vacuum energy does not gravitate.

Physicists treat this as a real, unsolved problem, not an apologist's invention. As quoted in Barnes's review: Burgess and Moore call it “arguably the most severe theoretical problem in high-energy physics today”; Bousso wrote in 2008 that “no concrete, viable theory predicting ρ_Λ = 0 was known by 1998 ... and none has been found since”; and Vilenkin judged that “the only viable resolution ... is provided by the anthropic approach”, that is, by a multiverse.¹ Note that last one: the physicists who take the problem most seriously mostly reach for a multiverse, not a designer.

The other dials, graded honestly

Popular lists contain dozens of “finely tuned constants”. When we apply the rules above, most shrink. This is the null space at work, and it decides where the argument can stand.

What survives. Two kinds of case hold up under multi-parameter variation with a physically motivated reference range: the cosmological constant (with the closely related inflaton energy) and the Higgs and quark-mass hierarchy. The nuclear cases have softened since 2009 thanks to better calculations, and the gravity figure is a category error. Barnes, who defends fine-tuning, says stars are “a bit more hearty than people gave them credit for”.¹⁰

The analysis

The argument in four lines

After Collins and Barnes,^{23, 24} in the form a skeptic can check:

1. Physics. For one universe in which nothing links the vacuum's starting value to particle physics, the chance of a life-permitting value is f ≤ 10^−50.9 (10^−48.6 if Q varies too), for any prior that doesn't already aim at the cancellation.
2. Design. P(E | D) = d: a mind that aims at embodied life has reason to choose life-permitting values, but its aims are not something we can derive. We let d run from 0.01 to 1.
3. Naturalism. P(E | N) = w·s + (1 − w)·f, where w is the weight naturalism gives to something that would make E expected (a multiverse with selection, or a law that cancels the vacuum energy), and s ≈ 1 is the chance it delivers life-permitting values.
4. Likelihood. E favours D over N by LR = d / [w·s + (1 − w)·f]. This measures confirmation. It does not give the probability of God, and it does not need “impossible without design”, which the evidence does not support.

Why the choice of prior no longer decides the answer for Λ

A probability needs a measure: a rule for how likely each value would be before we look. Critics are right that this choice can do all the work, and for most constants it does. For ρ_Λ it does not, once the question is put the right way:

• Put the probability on the vacuum's starting value, before the electroweak drop. Today's value is that starting value minus the drop, so a life-permitting universe needs the starting value to land within a few thousand units of a number 10^54.6 units away.
• Take any prior over the starting value with a single peak. If the peak is at or below zero, the chance of landing in that narrow band is at most its width divided by its distance: 5,100 / 10^54.6 ≈ 10^−50.9. That holds for a flat prior of any width and for a logarithmic one alike. Only a prior peaked within about 10% of the drop escapes, and such a prior already encodes the link to particle physics that “chance” denies.
• So the one assumption left is a principle: under one-universe chance, our expectation about the starting value does not anticipate the cancellation. It is a principle, not physics, and a critic can decline to hold any expectation at all (see the objections).

The computation

bound = window width / distance of the prior's peak from the band
      = 5,100 / 10^54.63 = 10^-50.9          (peak at or below zero)
general form: 10^-50.9 x D / |D - m|        (peak at m; D = the drop)

Now the honest test of the old method: change the measure on today's value itself. On a logarithmic measure over the size of today's vacuum energy, you need a floor (a smallest value considered), and the floor is arbitrary. As it drops, the “fine-tuning” evaporates:

That is why the question must be put on the starting value, not on today's value. Today's value is defined by subtracting the drop, so a prior that is spread evenly on its logarithm quietly assumes the cancellation it was meant to test. On the starting value, a logarithmic prior peaks at zero and gives 10⁻⁵⁷ or less.

For the other dials the measure is not a detail. For gravity or α, which cannot be zero, a log measure arguably is right,¹⁷ and that is why the gravity case shrinks (about 10⁻³⁸ on a linear axis, about 1 in 4 on a log axis); the Higgs value swings from 10⁻³² on a linear axis to between 0.2 and 0.8 on a log axis. That is why they stay out of the headline. One prior applied to every constant is a mistake.

The rivals: why the 1 in 10⁵¹ drops out

Chance is the weakest form of naturalism, and it fails. But naturalism is not committed to chance. It can appeal to a multiverse with selection or to a law that does the cancelling. Put any weight w on those, and the formula above collapses to LR ≈ d / w. The 10⁻⁵¹ term no longer matters at all.

And between design and those selectors, this evidence is silent. A cancelling law or a multiverse with selection makes a life-permitting universe certain; a designer who aims at life makes it likely. So what remains is a ratio of two judgments: how likely a designer is to aim at embodied minds (d), and how much weight a selector deserves (w). Neither can be derived. The designer's aims are as underived as the multiverse's measure, and we now hold both to the same standard.

That is the structural result of this page. The “1 in 10⁵¹” measures how badly chance fails. It does not measure how strongly design beats naturalism. The roll-up uses the second number, never the first.

Try your own numbers. The third slider shows that the physics inputs (the bound, or the first method's cut-off values) barely move the answer once w is above zero.

Why more tuned constants don't multiply

The quark masses, the Higgs value and Q all make chance (N0) even worse. They barely touch the multiverse (N2), because one landscape varies every constant at once. They do penalise a “deeper law” (N1), which would have to land every dial in its window. So extra constants raise the ratio against N1 but not against N2, and the bound LR ≲ d/w stays. That is why we never multiply the rows of the parameter table.

Where in the window does Λ sit?

A multiverse makes a further prediction a designer does not: how small a value a typical observer should see. Barnes and colleagues tested it with the EAGLE simulations. The probability that an observer sees Λ as small as ours depends on which measure over the multiverse you choose:

In their words, “the freedom to choose a measure is simply the freedom to choose predictions ad hoc”.² Our first version said that, since design makes no sharp prediction here, this datum favours design by up to 50 times. That was a mistake. “No sharp prediction” is not a probability of 1. A design hypothesis silent about where Λ sits still has to be modelled somehow, and every way of modelling it also puts our value low in the window by chance:

So the datum does not favour design. Under the causal measures it leans toward the multiverse, by a factor anywhere from 1 to about 280 depending on how one models a silent designer, which is the same kind of arbitrariness the measure problem charges the multiverse with. Barnes and colleagues' own summary, that structure formation “does not seem to be a sharp enough function of Lambda to explain its observed value alone”,² tells against the mass-weighted measure, not the causal ones. We add nothing to the ratio either way. Source: calc, section 9.

The result, with its uncertainty

Why these presets? For d = 0.3: a mind has some reason to make other minds, but we cannot derive its aims, and the universe is mostly hostile to life. For w = 0.03: the selectors are live candidates on several routes (a landscape, proposed cancelling mechanisms, a law-set that simply contains the cancellation), but none is established, the landscape's measure is underived, and its Boltzmann-brain conditions are open. Our first version set d = 0.5 and w = 0.01, discounting the multiverse for every unproven step and the designer for none; that is why the central value fell from 50 to 10, and why the high end, which needed a 1-in-10,000 weight on every selector together, fell from 10⁴ to 10³. The low end below 1 is deliberate: a selector makes a life-permitting universe certain, an aim makes it only likely. These are judgments, and we label them as such.

Objections

What this shows, and what it doesn't

What we don't claim

• Any compound “grand total” (10⁻¹⁴⁰ to 10⁻⁶²⁷, “574 bits”, “10²⁶² in favour of design”). They multiply dependent parameters, use unsourced windows and an inconsistent prior.
• “The cosmological constant is fine-tuned to 1 part in 10¹²⁰”, or “change it by one part in 10¹²⁰ and the universe flies apart”. The 10¹²⁰ is a mismatch that needs a Planck cut-off, not a tolerance. What holds: the vacuum energy must cancel an electroweak drop of about 10⁵⁵ times its present value, and for one universe by uncorrelated chance that is at most about 1 in 10⁵¹.
• “10⁻⁴⁰ or less, whatever the cut-off” (our own first headline). It treated the QCD term as the target; the cut-off-free bound replaces it.
• “P(fine-tuning | naturalism) is 1 in 10⁵¹.” That is the probability for one universe with nothing linking the vacuum to particle physics. Under a selecting law or a multiverse with selection it is about 1, so naturalism as a whole has no single number.
• Penrose's number as 10¹²⁰, or as evidence that Penrose infers God, or as the odds of our start. His figure is 1 in 10¹⁰¹²³, for a closed universe; he explains it by a law, and its probability measure is disputed. With a positive Λ the comparable figure is about 1 in 10¹⁰¹²², and it measures distance from equilibrium, not how special the start was.
• “The information needed to specify the universe exceeds its capacity, so its source is outside it.” That “33 times” result read a base-10 count as bits, and compared a figure with itself: picking one state never needs more bits than the space of states holds.
• “Gravity is fine-tuned to 1 part in 10⁴⁰” (or 10⁶⁰, or 10³⁴). Gravity could be about 10⁸ times stronger with stars still lasting millions of years. 10³⁶–10⁴⁰ is the size of a force ratio, not a window; 10⁶⁰ is the flatness figure; 10³⁴ depends on picking the strong force as the top of a linear range.
• Analogies that misstate the size. “A ruler across the universe adjusted by less than an atom” is about 1 part in 10³⁶, not 10¹²⁰; a chart of “0.0001% vs 99.9999%” shows 1 in 10⁶.
• “A 2% stronger strong force would turn all hydrogen to helium.” About 50% is needed, and diproton-burning stars still work.
• Rees's “ε = 0.007: at 0.006 only hydrogen, at 0.008 no hydrogen” as settled fact. Both ends are contested (unbound deuterium; the diproton).
• “The Hoyle resonance must be within 0.5%.” Lattice calculations give 2–3% as unlikely to be detrimental, and Adams gives about 1 in 4.
• “Flatness is tuned to 1 in 10⁵⁵–10⁶⁰” as a separate dial. Inflation addresses it, so it can't be counted alongside Q and Λ.
• “Matter–antimatter 1 in 10⁹” or “Q = 10⁻⁵” as tunings. Those are magnitudes, not windows.
• “The weak force is finely tuned.” A universe without it is arguably viable.
• “The speed of light is fine-tuned.” It has units, so its value depends on the units chosen.
• “Earth's odds are 1 in 10²⁸² or 10³⁹⁰, so Earth needs God.” Those are per-planet products of dependent, unmeasured factors, and Kepler finds roughly 0.4–0.6 rocky habitable-zone planets per Sun-like star.
• “Fine-tuning proves God”, “impossible by chance, so design”, “naturalism is falsified”. The argument is a likelihood comparison, strong only against chance, and eliminating chance leaves the other selectors standing.
• “Probability zero means impossible”, or “an infinite range only strengthens the argument”. Measure-zero events happen, and on an infinite range the probability is undefined, not zero.
• Unsourced totals such as “≪ 10⁻¹⁵⁰”, “10⁻¹⁸⁰”, “860 bits” or “10⁻⁶¹⁷”. No derivation exists; the last one's own inputs give 10⁻⁴³⁶, and over half of those rest on untraceable figures.
• “Any prior other than a flat one costs at least as much information as it saves.” False as stated; the true bound is weaker and applies to design as much as to its rivals.
• “Where Λ sits in its window favours design.” It does not; under the causal measures it leans, modestly and depending on modelling, toward a multiverse.
• “The generator must be tuned too, so the multiverse is refuted.” What is proved is that the specification is never smaller upstairs. As a refutation it needs a further premise, and it binds a designer's choice in the same way.
• “A multiverse needs infinitely many universes, which is incoherent.” A finite ensemble suffices: about 10⁵¹ domains.
• “Our universe has been shown to be optimal”, or “the constants have been derived from a deeper field”. No such calculation exists; the versions we have seen were fits to our own measured values, presented as derivations.
• “Design predicts that ever more fine-tunings will be found.” No such trend is shown, and several famous cases have softened.
• “Constants agree across regions that were never in contact, so something outside coordinated them.” Constants are parameters of laws, the same everywhere by definition.
• “Hossenfelder exempts the anthropic argument from her critique.” She sets it aside as a different logic; her objection to unexplained distributions still applies.
• “Fine-tuning points to the God of the Bible.” It does not discriminate among theisms.
• “The universe is optimally tuned for life.” Adams argues it could be more habitable.
• Famous quotes we haven't checked against the original. We don't use Hoyle's “superintellect” line or Dyson's “knew we were coming” (which is Dyson's, from Disturbing the Universe, 1979, though earlier versions of this argument credited it to Davies or Gonzalez), or lines attributed to Hugh Ross in second-hand summaries.
• The skeptic's overclaim: “supersymmetry or quantum gravity will obviously solve Λ.” Supersymmetry still leaves about 1 part in 10⁵⁰, and the problem appears well below the Planck scale.

Sources

1. Barnes, L. A. (2012). "The Fine-Tuning of the Universe for Intelligent Life." Publications of the Astronomical Society of Australia 29(4): 529-564. arXiv:1112.4647. Read: sections 4-6 and Appendix B. Also the source for the physicists quoted from it (Burgess and Moore; Bousso 2008; Vilenkin 2010; Penrose on roominess) and for studies marked “via Barnes”.
2. Barnes, L. A., Elahi, P. J., Salcido, J., Bower, R. G., Lewis, G. F., Theuns, T., Schaller, M., Crain, R. A. & Schaye, J. (2018). "Galaxy formation efficiency and the multiverse explanation of the cosmological constant with EAGLE simulations." MNRAS 477: 3727. arXiv:1801.08781. Read: text and Table 4.
3. Weinberg, S. (1987). "Anthropic bound on the cosmological constant." Physical Review Letters 59: 2607. doi:10.1103/PhysRevLett.59.2607. Not read in full; the bound is as reported in reviews.
4. Adams, F. C. (2019). "The degree of fine-tuning in our universe, and others." Physics Reports 807: 1-111. arXiv:1902.03928. Read: abstract, sections 1-2, part of section 7.
5. Martin, J. (2012). "Everything you always wanted to know about the cosmological constant problem (but were afraid to ask)." Comptes Rendus Physique 13: 566-665. arXiv:1205.3365. Read: section 3, eqs. 57-59.
6. Polchinski, J. (2006). "The cosmological constant and the string landscape." arXiv:hep-th/0603249. Read: sections 1.1-1.2.
7. Agrawal, V., Barr, S. M., Donoghue, J. F. & Seckel, D. (1998). "The anthropic principle and the mass scale of the Standard Model." Physical Review D 57: 5480. arXiv:hep-ph/9707380. Read: abstract.
8. MacDonald, J. & Mullan, D. J. (2009). "Big bang nucleosynthesis: the strong nuclear force meets the weak anthropic principle." Physical Review D 80: 043507. arXiv:0904.1807. Read: abstract.
9. Barnes, L. A. (2015). "Binding the diproton in stars: anthropic limits on the strength of gravity." JCAP 2015(12): 050. arXiv:1512.06090. Read: abstract.
10. Popkin, G. (2017, May 12). "Reports of universe's fine-tuning may be exaggerated." Inside Science (AIP). aip.org. Read. Also our source for Adams & Grohs (2017) on stars with unstable deuterium.
11. Barnes, L. A. & Lewis, G. F. (2017). "Producing the deuteron in stars: anthropic limits on fundamental constants." JCAP 2017(07): 036. arXiv:1703.07161. Read: abstract.
12. Epelbaum, E., Krebs, H., Lähde, T. A., Lee, D. & Meißner, U.-G. (2013). "Dependence of the triple-alpha process on the fundamental constants of nature." European Physical Journal A 49: 82. arXiv:1303.4856. Read: abstract.
13. Harnik, R., Kribs, G. D. & Perez, G. (2006). "A universe without weak interactions." Physical Review D 74: 035006. arXiv:hep-ph/0604027. Not read directly; reported via Barnes (2012).
14. Grohs, E., Howe, A. R. & Adams, F. C. (2018). "Universes without the weak force: astrophysical processes with stable neutrons." Physical Review D 97: 043003. Not read directly.
15. Tegmark, M., Aguirre, A., Rees, M. J. & Wilczek, F. (2006). "Dimensionless constants, cosmology and other dark matters." Physical Review D 73: 023505. arXiv:astro-ph/0511774. Read: abstract; the 8 constraints on 7 parameters (their Table 4) via Barnes (2012).
16. Epelbaum, E., Krebs, H., Lähde, T. A., Lee, D. & Meißner, U.-G. (2013). "Viability of carbon-based life as a function of the light quark mass." Physical Review Letters 110: 112502. arXiv:1212.4181. Read: abstract.
17. Adams, F. C. (2008). "Stars in other universes: stellar structure with different fundamental constants." JCAP 2008(08): 010. arXiv:0807.3697. Read: abstract.
18. Penrose, R. (1989). The Emperor's New Mind. Oxford University Press; Penrose, R. (2004). The Road to Reality. Jonathan Cape, pp. 726-732, 762-765. Not read directly; the figure is reported via Barnes (2012) and Goldstein (2001, arXiv:cond-mat/0105242).
19. Egan, C. A. & Lineweaver, C. H. (2010). "A larger estimate of the entropy of the universe." Astrophysical Journal 710: 1825-1834. arXiv:0909.3983. Not read directly here; the horizon entropy of order 10¹²².
20. Earman, J. (2006). "The ‘past hypothesis’: not even false." Studies in History and Philosophy of Modern Physics 37: 399-430. Not read directly.
21. Bryson, S. et al. (2021). "The occurrence of rocky habitable-zone planets around solar-like stars from Kepler data." Astronomical Journal 161: 36. arXiv:2010.14812. Read: abstract.
22. Ward, P. & Brownlee, D. (2000). Rare Earth: Why Complex Life Is Uncommon in the Universe. Copernicus. Not read directly.
23. Collins, R. (2009). "The teleological argument: an exploration of the fine-tuning of the universe." In W. L. Craig & J. P. Moreland (eds.), The Blackwell Companion to Natural Theology, pp. 202-281. Wiley-Blackwell. Not read directly; content via the Stanford Encyclopedia and Barnes (2012).
24. Barnes, L. A. (2019). "A reasonable little question: a formulation of the fine-tuning argument." Ergo 6(42). philarchive.org. Not read in full.
25. Susskind, L. (2003). "The anthropic landscape of string theory." arXiv:hep-th/0302219. Read: abstract.
26. Martel, H., Shapiro, P. R. & Weinberg, S. (1998). "Likely values of the cosmological constant." Astrophysical Journal 492: 29. arXiv:astro-ph/9701099. Read: abstract.
27. Carroll, S. M. (2017). "Why Boltzmann brains are bad." arXiv:1702.00850; in Current Controversies in Philosophy of Science (Routledge, 2020). Read: abstract.
28. Hacking, I. (1987). "The inverse gambler's fallacy: the argument from design. The anthropic principle applied to Wheeler universes." Mind 96(383): 331-340. Not read directly.
29. White, R. (2000). "Fine-tuning and multiple universes." Noûs 34(2): 260-276. Not read directly.
30. McGrath, P. J. (1988). "The inverse gambler's fallacy and cosmology: a reply to Hacking." Mind 97: 265-268; Leslie, J. (1988). "No inverse gambler's fallacy in cosmology." Mind 97: 269-272. Not read directly.
31. Friederich, S. (2017; revised). "Fine-Tuning." Stanford Encyclopedia of Philosophy. plato.stanford.edu. Read: sections 1-5. Our source for Epstein (2017), Isaacs, Hawthorne & Russell (2022), Draper (2020), Barrett & Sober (2020) and Howson on old evidence.
32. Sober, E. (2003). "The design argument." In N. Manson (ed.), God and Design. Routledge. Not read directly; via the Stanford Encyclopedia.
33. Leslie, J. (1989). Universes. Routledge. Not read directly.
34. Weisberg, J. (2005). "Firing squads and fine-tuning: Sober on the design argument." British Journal for the Philosophy of Science 56: 809-821. Not read directly.
35. McGrew, T., McGrew, L. & Vestrup, E. (2001). "Probabilities and the fine-tuning argument: a sceptical view." Mind 110(440): 1027-1038. Not read directly.
36. Colyvan, M., Garfield, J. L. & Priest, G. (2005). "Problems with the argument from fine tuning." Synthese 145: 325-338. Not read directly.
37. Pruss, A. (2005). "Fine-tuning and coarse-tuning, normalizability, and probabilistic reasoning." Philosophia Christi 7(2): 405-423. Not read directly.
38. Koperski, J. (2005). "Should we care about fine-tuning?" British Journal for the Philosophy of Science 56: 303-319. Not read directly; via the Stanford Encyclopedia.
39. Norton, J. D. (2010). "Cosmic confusions: not supporting versus supporting not." Philosophy of Science 77(4): 501-523. pitt.edu. Read: abstract.
40. Hossenfelder, S. (2019). "Screams for explanation: finetuning and naturalness in the foundations of physics." Synthese. arXiv:1801.02176. Read: sections 2.2-2.3, 5.1.
41. Ellis, G. F. R., van Elst, H., Murugan, J. & Uzan, J.-P. (2011). "On the trace-free Einstein equations as a viable alternative to general relativity." Classical and Quantum Gravity 28: 225007. arXiv:1008.1196. Abstract read by an earlier reviewer, not by us.
42. Kaloper, N. & Padilla, A. (2014). "Sequestering the Standard Model vacuum energy." Physical Review Letters 112: 091304. arXiv:1309.6562. Abstract read by an earlier reviewer, not by us.
43. Stenger, V. J. (2011). The Fallacy of Fine-Tuning: Why the Universe Is Not Designed for Us. Prometheus. Not read directly; via Barnes (2012).
44. Dawkins, R. (2006). The God Delusion, ch. 4. Bantam. Not read directly.
45. Narveson, J. (2003). "God by design?" In N. Manson (ed.), God and Design. Routledge. Not read directly; via the Stanford Encyclopedia.
46. Monton, B. (2006). "God, fine-tuning, and the problem of old evidence." British Journal for the Philosophy of Science 57: 405-424; Glymour, C. (1980). Theory and Evidence. Princeton University Press. Not read directly; via the Stanford Encyclopedia.
47. Dicke, R. H. (1961). "Dirac's cosmology and Mach's principle." Nature 192: 440-441. Not read directly.