In epistemology, we often think of the things we believe as discrete propositions. For instance, you may believe that there is a computer screen in front of you. But how is this belief justified? One way of justifying a belief is by offering a reason, which can itself also be a proposition. For this next proposition, we can then ask how it is justified and so on. The regress problem asks the following question: if any of the things we believe are justified, then what is the structure of that justification? Does the justification question not just keep getting passed backward forever with reasons for reasons for reasons?

To illustrate the common responses to the regress problem, imagine playing a game of telephone with a (potentially infinite) group of players. Telephone is played by secretly passing a message between players and comparing the message received at the end to messages received earlier in the game. Suppose you receive the message $M$ and wish to follow the message backward to compare the message you received to the original. You proceed by following the path of message passers backward and asking for the content of the message they had received. We can call the first message back $M_1$, the next message back $M_2$, and so forth.

Below, we consider three common responses to the regress problem using this example. For simplicity, we will assume that a reason given to support a proposition is always a good reason and that a message is always passed from one player to the next iteratively; there are never two incoming messages at the same time. (We will drop this assumption in part II.) This results in a chain of messages of the following form: $M \leftarrow M_1 \leftarrow M_2 \leftarrow M_3 \leftarrow \cdots$.

Foundationalism

Suppose after tracing our message backward, the chain ends after $n$ steps at the original message $M_n$, much like with the usual game of telephone. We can compare the received message $M$ with original message $M_n$ to see how much the message was distorted through the passing chain. If $M$ and $M_n$ contain similar information then the playing group succeeds; otherwise, they fail. Notice that unless the game could have stopped at any point the comparison that really matters is between $M$ and $M_n$ and not the comparisons with intermediate messages.

Similarly, for the foundationalist, the ultimate justification for our beliefs rests on these basic or foundational beliefs. Just as $M_n$ is where the message originates, basic beliefs are in some sense self-justified. Candidates for these basic beliefs include sense perceptions and naive seemings. For example, I am experiencing an itch or seeing a red triangle, or it seems that I have hands. The primary difficulty with this position is to give a clear account of basic beliefs: what they are and what are their epistemological properties. Foundationalism has been the dominant response to the regress problem until the second half of the 20th century, when coherentism gained popularity.

Coherentism

Back in the game of telephone, imagine that after $n$ steps we reach a message that we’ve seen before. In this odd situation, we appear to be stuck in a loop as tracing the message onward would lead us to traverse the same sequence of messages again and again. How might we go about judging if the game has been played successfully or not? One method is to compare $M$ with all messages in the loop to see if they jointly contain similar information. Success results only in the case where all of the selected messages contain similar information.

With respect to the justification of beliefs, the coherentist argues that justification emerges from the structure of a system of beliefs rather than ultimately resting on beliefs that are in some way self-justified. Consider a case of how this may work in practice, where we receive independent testimony of several witnesses to an event. Individually, we may not have a reason to believe the content of their testimony; however, if each witness reports a similar sequence of events, then belief in their individual reports may be more justified. We can say that these testimonies mutually support one another. On this account, this mutual support or coherence is what drives the justification of our beliefs.

Many approaches in the contemporary literature combine elements from coherentism with a weak form of foundationalism, taking the best of both approaches.

Infinitism

In this twist on game of telephone, suppose tracing the message backward is never-ending and we never repeat messages. This variant of telephone requires either infinitely many players or, at least, infinitely many potential players. As with coherentism, we find ourselves in a situation where judging success is unclear. One method is to compare $M$ to $M_n$ as $n$ grows ever larger. Suppose that each time we are prompted to judge the success of the game we incrementally check a larger $n$ and find that $M$ and $M_n$ contain similar information. In that case, we can say, at least provisionally, that the game has been played successfully.

In contrast to foundationalism and coherentism, the infinitist holds that the infinite regress is not problematic. One popular position defending infinitism argues that to be justified in a belief one need only be able to give an additional reason when prompted, since no finite or human-like agent can offer infinitely many beliefs, as with the infinite variant of the telephone game above. Infinitism is not without its difficulties. At the least, one needs to give an account of what it means to both have access to and always be able to give another reason.

Arguments for this brand of infinitism were introduced by Peter Klein and others in the late 20th Century.

Conclusion

The regress problem is far from settled. In the nearly two thousand years since its introduction, great strides have been made in understanding the structure of justification for our beliefs. Currently, some form of foundationalism is the dominant position, coherentism still has its defenders but has been in recent decline, and infinitism has started to be considered as a serious position during the last twenty years. Additionally, over the past decade, a new mathematical and probabilistic approach has emerged to recast the regress problem in more formal and interdisciplinary terms.

In Part II of this article (coming soon), we will extend the analogy of the telephone game to allow for more complex forms of message passing. In turn, this will allow for a more robust characteriztion of the structure of epistemic justification, especially for both both finite and infinite varieties of coherentism.

References

The material below is open access and available to all.

On foundationalism for justification in epistemology:
Ali Hasan and Richard Fumerton. “Foundationalist Theories of Epistemic Justification”, The Stanford Encyclopedia of Philosophy (Winter 2016 Edition), Edward N. Zalta (ed.), https://plato.stanford.edu/archives/win2016/entries/justep-foundational

On coherentism for justification in epistemology:
Erik Olsson. “Coherentist Theories of Epistemic Justification”, The Stanford Encyclopedia of Philosophy (Spring 2017 Edition), Edward N. Zalta (ed.), https://plato.stanford.edu/archives/spr2017/entries/justep-coherence/

On infinitism for justification in epistemology:
Peter Klein and John Turri. “Infinitism in Epistemology”, Internet Encyclopedia of Philosophy, https://www.iep.utm.edu/inf-epis/

On the general form of regress arguments in philosophy:
Cameron Ross. “Infinite Regress Arguments”, The Stanford Encyclopedia of Philosophy (Fall 2018 Edition), Edward N. Zalta (ed.), https://plato.stanford.edu/archives/fall2018/entries/infinite-regress/

On the mathematical approach to the regress problem:
David Atkinson and Jeanne Peijnenburg. 2017. Fading Foundations: Probability and the Regress Problem. Springer Open. https://doi.org/10.1007/978-3-319-58295-5