In the famous 1999 film The Matrix, a world is shown in which humans have been unknowingly trapped inside a simulated reality. Although people in the Matrix appear to live normally, they are completely unaware that their perception has been falsified by intelligent machines. Thought experiments of this nature have been popular throughout history, with Plato’s ancient allegory of the cave illustrating a plausible scenario in which people manipulated the reality prisoners believed in using shadow images.
The fear of being deceived is a ubiquitous manifestation of the fear of the unknown. Unlike predictable, clearly perceived threats, being fooled puts the very instruments which you use to detect danger into doubt. In Descartes’s 1641 book Meditations on First Philosophy, he tries to illustrate the most extreme degree of being deceived.
The first section is titled “What Can Be Called into Doubt”, and it begins with Descartes recognizing that he’s believed many falsehoods since he was young. Since his knowledge as an adult relies on “shaky foundations” built as a child, Descartes decides that in order to build something stable in his scientific studies, he must begin from the bottom up and find what he can be certain of.
The bootstrapping problem
The world of software has a similar problem. Let’s say you’re programming a tool, and this tool needs to transmit data over the internet. In order to do this securely, you’ll need the OpenSSL library. However, the OpenSSL library needs gcc, a compiler that turns C/C++ into machine code. Furthermore, gcc itself needs a C standard library like glibc which implements functions like malloc or printf.
Although your tool’s source code may be publicly available, what if the specific build of OpenSSL that your tool depends on has a backdoor written in it? If your code calls functions from OpenSSL, but the library itself has been compromised, you can’t trust the tool. The most straightforward solution you could propose is simply building OpenSSL from source. If you download clean code, then you could build your tool against this locally-built OpenSSL.
However, we still aren’t safe. Remember that in order to build OpenSSL from source, we need a C/C++ compiler like gcc. Imagine our compiler binary has implemented a rule along the lines of “If I am being used to compile OpenSSL, add malicious functions into the resulting binary.” This problem was famously stated in a 1984 lecture by Ken Thompson called “Reflections on Trusting Trust”. Just as Descartes doubts the knowledge he gained in his childhood, Thompson scrutinizes the tools used to compile tools.
Thompson’s lecture concludes that you can’t trust software not written by yourself. No matter how much you inspect source code, well-placed bugs can be placed one level lower, whether that be in an assembler or hardware microcode. Therefore, trusting software is ultimately an act of faith.
In Descartes’s “What can be called into doubt”, he imagines a worst-case scenario: an evil demon with extreme power determined to make him believe false things. The demon can fabricate what he sees and manipulate how he perceives his body. His hands, eyes, flesh, and even his senses could be tricks. It can falsify “fundamental truths”, like making Descartes believe that squares have four sides or two plus three making five despite these being ultimately untrue.
Descartes’s Seed
In the second section of Meditations, titled “The Nature of the Human Mind”, he establishes that the world of ideas in his mind represented things that were often external to it. Since it’s possible for ideas to be inaccurate representations, Descartes is led to a search for a kind of bridge that connects the world to the mind. His first step is to find an idea that is certainly accurate.
Assuming that all which the demon can manipulate is false, Descartes concludes there is one truth he can be sure of: his own existence. This is because for Descartes to be deceived by the demon, he must exist. The famous phrase cogito, ergo sum (which translates to “I think, therefore I am”) originates from this thought experiment.
In the same section, Descartes searches for what “I” means. He defines himself to be “a thinking thing”, and “a thinking thing” to be “something that doubts, understands, affirms, denies, wills, refuses, and also senses and has mental images”. This definition has a basis in pure rationality, which gives it certainty. While the demon can falsify Descartes’s vision to make him believe he is staring at an object when it does not exist in reality, the statement “it seems I am looking at this object” cannot be falsified in the same way.
Despite understanding that the things he rationally understands in his mind are known more certainly than that which he physically senses, Descartes still feels that concrete perception is more clear and distinct. After all, the source of truth for many facts a person believes is simply what they have witnessed for themselves. To correct this, Descartes uses the example of wax from a honeycomb. When solid, it smells like flowers, tastes sweet, and creates a sound when tapped. When the wax is melted, the smell evaporates, the taste disappears, and is silent when tapped. Despite having different physical attributes, we still understand it to be the same wax because we clearly and distinctly perceive its essence through reason alone.
Tensed truth values
While Descartes continues Meditations by exploring the existence of a God, it could be worth returning to his worst-case scenario of deception and scrutinizing if it truly is the extent to which one can be fooled. The malevolent demon is a natural extrapolation of real-life examples of deception, but it still follows the rules of conceivable reality. What if the true extent of deception were simply incomprehensible to the human mind?
Let’s take a closer look at the meaning of “I exist”, which Descartes is certain of. In the original Latin text of Meditation II, it is phrased:
“‘Ego sum, ego existo’, quoties a me profertur, vel mente concipitur, necessario esse verum.”
This is a smaller claim than the popularized “Cogito, ergo sum”, as it translates to “‘I am, I exist’, whenever it is spoken by me, or conceived in my mind, is necessarily true.” The statement appears self-evident, as the declaration of his existence is time-dependent and still holds true notwithstanding a case where he oscillates in and out of existence every second, because he’d only be able to speak or conceive of his existence while he exists.
However, if a statement’s truth value can be time-dependent, when is it evaluated? Take the word “existo”. In Latin, the ending “-o” indicates that the verb is first person and present tense, which gives it the translation of “I exist”. But if this is only true when Descartes speaks it, does this hold over the duration of the entire sentence?
Of course, since Descartes must exist in order to speak, this is not an issue for his declaration of existence. However, it does raise an interesting question about what exactly truth values track in the real world. For example, imagine sitting in front of a campfire and saying “The fire is burning right now.” The twist is that the fire extinguishes as your mouth forms the word “now”. The verity then becomes unclear.
This is a problem for any ambiguous statement. If you were to evaluate the truth value of the statement “There are bats inside the cave” but the cave contained wood clubs used for a sport instead of the winged mammal, is the proposition definitively true or false? Of course, one may argue that there are clearly statements that have no truth value at all, such as a nonsensical string of letters, but does there exist an objective, reproducible way to determine if a statement is truth-apt?
There exist views that take a departure from the common conception of every proposition being true or false, such as Graham Priest’s dialetheism, which suggests that there are statements which can be both true and false, as well as true statements whose negations are also true. Of course, since this is a thesis about truth, is dialetheism itself true?
I haven’t looked very deeply into dialetheism beyond reading Priest’s book Logic: A Very Short Introduction, but many debates regarding its validity appear to center around its ability to capture our social capacity of disagreement (a true statement whose negation is also true means that two people can “disagree” still accept each other’s statement as true), suggesting that our conventions of communication play a role in assessing what “truth” means.
Descartes’s dependence on language
In Meditations on First Philosophy, Descartes stated that his goal was to doubt everything that stood on uncertain epistemological grounding and find what he was truly certain of. However, in doing so, he used language, and if language comes with presumptions, his certainty may stand on top of those.
Additionally, as shown by the existence of dialetheism and the conversation around different conceptions of truth, logic may also be dependent on language. As a straightforward thought experiment demonstrating what this might look like, imagine a civilization of humans that can communicate verbally. Each member of this group has never knowingly deceived or been deceived by any other member, and they have never created a hypothesis that was later found to be proven wrong. Their language might entirely lack analogs to truth values.
Since Descartes wrote in Latin, a descendant of Proto-Indo-European, he might be facing the same problem. If the language and frameworks he thought in prohibited him from doubting propositional logic or something more foundational, it’s possible that cogito, ergo sum is known certainly in the same way that the previously mentioned civilization “knows” information spoken by any member.
The word “certainty” in Thompson’s case bottoms out at a certain understanding of physics. Although a dedicated actor could implement an attack through hardware microcode, a being which knows the state of every particle in the universe could be “certain” that no vulnerabilities exist if he has manually inspected every single transistor and logic gate, verifying that a program does as he expects. (Thompson’s conclusion of trusting software being faith-based is practically valid because we don’t know of such a being, at least not one that works in the software industry.)
With Descartes, “certainty” no longer depends on sensory information, as it’s rationally conceivable for them to be deceived. Instead, only pure reason is trusted because it’s not rationally conceivable for valid logic to be false. However, the divergence of what “rationally conceivable” means complicates the existence of that floor. If a dialetheist disagrees with the mutual exclusivity of truth values, examining an argument to be valid no longer does the same thing as the logic gate verification.
Perhaps every logical framework has its own corresponding certainty, but the “truth” of these certainties wouldn’t be mind-independent. The only way for there to be an undoubtable universal truth is if “truth” means the same thing regardless of who speaks it. Any attempt to resolve this fragmentation by declaring either that there is an ultimate conception of truth or by declaring that none exists requires a conception of truth. Ultimately, it appears that when one recursively doubts language and logic, incoherence is inevitable.