The Epistemic Problem

  • Sjoerd D. Zwart
Part of the Synthese Library book series (SYLI, volume 307)


In the first chapter, I mentioned the difference between the semantic (1) and the epistemic (2) problem of approach to the truth, and I have dedicated the first three chapters of this book to the first problem. In the present chapter, we shall study a more practical subject, viz. the second problem and three answers to it. Let me formulate the two questions explicitly:
  1. (1)

    The semantical problem: “What do we mean if we claim that the theory ψ is closer to the truth than φ?”

  2. (2)

    The epistemic problem: “On what evidence are we to believe that the theory ψ is closer to the truth than φ?”



Preference Order Impossible World Success Rule Empirical Success Crucial Experiment 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Note the subtle difference. The content approach considers the true consequences of the theories; the likeness rule is based on the true consequences of all non-falsified constituents.Google Scholar
  2. 2.
    Kuipers (1992), p.326.Google Scholar
  3. 3.
    Irrevisable and irreversible rules may be defined as follows. Let p be a rule of theory-choice; let e be its preference ordering based on evidence e, Then p is irrevisable iff w e cp implies Ve’: e’ e and w e, cpGoogle Scholar
  4. irreversible iff W e cp implies b’e’: e’ e and cp é, yJGoogle Scholar
  5. Note that an irrevisable rule is functional, and a functional rule is irreversible.Google Scholar
  6. 4.
    Niiniluoto (1987), p.482. note 5.Google Scholar
  7. 5.
    Niiniluoto’s point of view regarding Laudan’s challenge can be found in Niiniluoto (1997).Google Scholar
  8. Ibid. sect.6.8.Google Scholar
  9. 7.
    See Niiniluoto (1997).Google Scholar
  10. 8.
    Popper (1963), p.234.Google Scholar
  11. 9.
    Niiniluoto (1987), p.264–265.Google Scholar
  12. 10.
    Popper (1963), Appendix IX, p. 288.Google Scholar
  13. 11.
    Popper (1963), p.235.Google Scholar
  14. 12.
    . Every reader, knowing more than the barest outlines of falsificationism, knows that my presentation is grossly simplified. For example, Lakatos (1970) sketches a gamut of falsificationist positions entirely neglected in our exposition. Taking Lakatos’s idea of scientists seriously using systematic strategies to refute theories, van Benthem suggests an interesting line of further research linking belief-revision and approach to the truth. We could classify the various revision strategies in the light of approach to the truth.Google Scholar
  15. 13.
    This procedure is well illustrated by a comparison of Popper. He compared the search for truth to the search for a “black cat in a dark room that might not even be there”, with thanks to A. Keupink who drew my attention to this fact.Google Scholar
  16. 14.
    The elements of Ix can also be interpreted as partial models that together constitute one big model of reality.Google Scholar
  17. 15.
    Kuipers (1996, p.86) calls S(t) a general fact since “scientists use to speak about a general fact.”Google Scholar
  18. 16.
    Kuipers (1987), p.96.Google Scholar
  19. 17.
    Kuipers (1995), p.365.Google Scholar
  20. 18.
    See Kuipers (1992), p.325.Google Scholar
  21. 19.
    Kuipers (1992), p.326.Google Scholar
  22. 20.
    See Kuipers (1998).Google Scholar
  23. 21.
    Kuipers (1992), p.308, last paragraph.Google Scholar
  24. R(t) I trivially obtains, because R(t) are situations in reality that are part of the intended applications.Google Scholar
  25. 23.
    Kuipers (1992), p. 301.Google Scholar
  26. 24.
    Kuipers (1992), p. 309.Google Scholar
  27. 25.
    See for similar objections Niiniluoto (1987), p.381.Google Scholar
  28. 26.
    Recall that if we consider a new R’ and S’ increase of R and decrease of S both represent increase of logical strength.Google Scholar
  29. 27.
    Kuipers (1996, p.87, p.97) changed his terminology but the principle remained the same. He uses the term “general facts” instead of “the strongest accepted law until time t”, and S(t) is substituted by ES(M, t): the explanatory successes of theory M at time t. It contains I as a subset. He gives “corrected versions of the laws of Galileo and Kepler” as examples of the general test implications that are supposed to be true.Google Scholar
  30. 28.
    As mentioned earlier, substitution of the average sum of the minimal distances for the sum of those distances would balance the situation.Google Scholar
  31. 29.
    Kuipers (1982), p. 357.Google Scholar
  32. 30.
    For more instances see Niiniluoto (1987), p.2–18.Google Scholar
  33. 31.
    Niiniluoto presents his answer to the epistemic problem in (1987), chap 7.Google Scholar
  34. 32.
    i. Ibid. p. 269.Google Scholar
  35. 33.
    i. Ibid p. 270.Google Scholar
  36. 34.
    Niiniluoto (1987), the seventh note of chapter 7.Google Scholar
  37. 35.
    See Niiniluoto (1987), p.270.Google Scholar
  38. 36.
    Niiniluoto (1987), p.274 formula (19).Google Scholar
  39. Ibid. Result (IV) of equation (18).Google Scholar
  40. Ibid p.273Google Scholar
  41. 39.
    Niiniluoto (1987), p.275.Google Scholar
  42. 40.
    For the information of the foregoing paragraph, see Niiniluoto (1987), p. 275275.Google Scholar
  43. 41.
    i. Ibid. p. 268–269.Google Scholar
  44. 42.
    . As Carnap ( 1962, p.523) explains, we could have made another choice. The mode or median are also candidates. The latter possibilities, however, are less satisfactory.Google Scholar
  45. 43.
    Niiniluoto (1987), p.269.Google Scholar
  46. 44.
    i. Ibid. p.269, 270.Google Scholar
  47. 45.
    Carnap (1962, chap IX) warns for this kind of problem.Google Scholar
  48. 46.
    Here ver(h3/e) = 1–184815y-0,404y’, and ver(h5/e) = 1–214815y-0,0742y’Google Scholar
  49. 47.
    Here ver(h4/e)= 1–216990y - 0,0508y’, and ver(h5/e) =1–221275y-0,0666y’Google Scholar
  50. 48.
    Niiniluoto (1987), section 12.5 especially p.426.Google Scholar
  51. 49.
    One consequence of the naive or content rule is that it does not decide between theories in the following specific situation. Suppose two theories explain the strongest law and respect all instantial data. In that circumstance, Popper would prefer the logically stronger theory, but the success rule 4.2 (p. 132) is indifferent. If it preferred the strongest theory, then it would cease to be functional for approaching the truth. Extension ofR could put the stronger theory at a disadvantage.Google Scholar
  52. 50.
    Niiniluoto (1987) p.269.Google Scholar
  53. 51.
    i. Ibid. chap. 12.5.Google Scholar
  54. 52.
    See also Bonilla (1996) and Kieseppä (1996).Google Scholar
  55. 53.
    See also Kuipers (1987).Google Scholar
  56. 54.
    Kuipers (1992), p.326.Google Scholar
  57. Since 1995, Kuipers presents his rule of theory choice in combination with the HD-method. He formulates his rule of success conditionally: “on the basis of comparative HD-testing, it appears that the theory Y will remain more successful than X”.Google Scholar
  58. 56.
    See e.g. Niiniluoto (1987), p.380–382.Google Scholar
  59. 57.
    From a private letter to Theo Kuipers dating from 1983.Google Scholar
  60. 58.
    See van Benthem (1996).Google Scholar
  61. 59.
    Preferential reasoning started with Shoham (1988); for conditional logic see Friedman and Halpern (1994). Gärdenfors and Rott (1995) provide an introduction on belief revision.Google Scholar

Copyright information

© Springer Science+Business Media Dordrecht 2001

Authors and Affiliations

  • Sjoerd D. Zwart
    • 1
  1. 1.Delft University of Technology and University of AmsterdamThe Netherlands

Personalised recommendations