THE
WESTERN INSTITUTE
FOR
ADVANCED STUDY
IS THERE A SUCH THING AS TRUTH?
Truth, and our ability to discover it - what do scientists themselves believe?
At the Western Institute for Advanced Study, we like to tackle the big questions. So this year, we asked our Board of Scientific Advisors to engage with a really deep question: Is there such a thing as truth? We then asked a critical follow-up question: Even if there is a such thing as truth, can we really know what's true? Surprisingly, the answers were quite diverse, but they converged in an interesting way.
​​
The first question was simply: Do you believe in truth? For example, do you believe there could be a comprehensive theoretical framework explaining our world (at the physical, biological, or social level) that remains constant over time and accommodates all perspectives and observations?
​
Richard DeMillo, who served as Chief Technology Officer at Hewlett Packard and is now the Charlotte B. and Roger C. Warren Chair in Computing at Georgia Institute of Technology, answered this question in the negative. He outright rejected the very concept of “an objective, observer-independent concept called truth” and found the question itself to be problematic, “in the sense that there is something called objective reality whose properties can be discerned with certainty.” To support his view on the matter, he gave both logical and empirical arguments: “We know, for example, that there are mathematical universes consistent with the Axiom of Choice and others with its negation. This point of view is distinct from, but consistent with, the probabilistic nature of measurement.” However, DeMillo noted the weight of validity: “Scientific argumentation is a dialog aimed at increasing confidence, that a chain of reasoning that leads from a set of propositions (laws, theorems, observations) believed to be free of contradictions does not lead to a contradiction. This resembles the Socratic rule: ‘Agree with me if I seem to speak the truth.’”
​
Other respondents built on this notion of seeking validity through consensus. Steven Zeichner, McLemore Birdsong Professor of Pediatric Medicine at University of Virginia, responded that he believed in truth – with a caveat. “It depends on what is meant by ‘all perspectives and observations’. Some people have perspectives that may not accord with ‘the truth’ as understood by others.” Karl Friston, Professor of Theoretical Neuroscience and Director of the Wellcome Centre for Human Neuroimaging at University College London, found his own theory of active inference to be useful in making sense of this observation: “[Truth] is a local phenomenon, unless the evidence — upon which inference is based — is shared among truth-seeking observers…. We are all observers seeking the best explanation for the observed world.”
Charles Simon, Founder of the Future AI Society and a long-time tech entrepreneur, answered: “Yes, our reality is explainable even if we cannot (yet) explain it. An alternative definition for truth is one of internal consistency. If we live in a simulation, for example, we potentially will not be able to tell because the simulation is completely consistent. This limits our concept of truth to the reality that we (or our instruments) can detect.”
​
So even if we live in a simulation, we can at least discover what's true within that simulation!
The second question of the survey was: Do you believe it is possible to discover what is true? For example, do you believe that our understanding of truth is imperfect, that truth itself is elusive, or that ‘truth’ is a merely local phenomenon, subject to shifting conditions and perspectives?
​
Bharat Ratra, Distinguished Professor of Physics at Kansas State University and a leader in the field of dynamical dark energy, gave a thought-provoking answer: “I believe in a reality that exists independent of us and that science proceeds by developing better and better approximations to this reality, based on theoretical models and experimental/observational tests. [However] I do not know how this process will play out… [If the question is] whether reality can evolve, I think the answer is yes. For example, the universe expands, and the expansion rate changes, and if dark energy and dark matter and baryonic matter eventually becomes subdominant to something our current measurements are not sensitive to, then the future universe could look very different compared to what it looks like now.”
​
Is science capable of keeping up with a changing universe? Eric Schlegel, Vaughn Family Professor of Physics and Astronomy at the University of Texas at San Antonio, helpfully rephrased the question: “Can physics be expounded to be comprehensive over all aspects of perspectives and observations? Many would argue 'no', but I am less certain that 'no' is the correct response. Physics has been and is tackling non-linear problems which means it could very well encompass all.” But he then added: “Truth, particularly from the view of science, is always incomplete. It is easy to deduce that unsupported objects fall toward the center of the Earth. It is more difficult to measure the rate at which they do so. It is even more difficult to discern that a gravitational redshift exists that must be corrected using general relativity. At the current time, we think we understand how the Universe works, but there are some inconsistencies in that understanding. Assuming more evidence accumulates that demonstrate that our current understanding is incomplete, then we will begin to see a path toward an improved understanding, hence, closer to the truth.” Maybe we keep inching toward the truth, but the full picture escapes us, and we struggle with incompleteness.
​
Suzanne Still, Professor of Computer Science at the University of Hawaii and a trailblazer in the emerging field of thermodynamic computation, also had a clear and thought-provoking answer to this question about whether we could reach the truth. Her answer was simply: “asymptotically”. This view was shared by Yasunori Nomura, Director of the Berkeley Center for Theoretical Physics and an Elected Fellow of the American Physical Society, who elaborated on this response: “We can uncover truth ‘asymptotically’. It is not clear if we can reach a ‘closed’ framework with which we can understand everything.” It’s worth noting, in this context, that Still’s seminal 2012 paper ‘The thermodynamics of prediction’ – along with a recent publication building on this work - formally demonstrate the thermodynamic limits to knowledge acquisition, in striking alignment to Godel’s incompleteness theorems. The case may be that we can never know the complete truth, even if it does exist.
​
Interestingly, Schlegel added a further insight, which corresponded enormously well with Friston’s and Still’s work on consistency and prediction in the context of computation: “Interestingly, at the moment this survey arrived in my inbox, I was attending an "AI" large language model (LLM) session and learned something about discursive networks. If a discursive network has a single source of information, most people (or even LLMs) will 'believe' whatever is generated by the information source. If there are even just two sources of information, the discursive networks [and people] follow the 'correct' information, rather than the inaccuracies. That's essentially the description of the scientific approach -- check and verify, then build upon what you've learned.”
​
This exercise has brought an important scientific challenge into clear focus.
​
Stuart Firestein, Chair of the Department of Biological Sciences at Columbia University and author of Ignorance: How It Drives Science, noted the following in his response: “Neils Bohr said the opposite of a fact is a falsehood, but the opposite of a profound truth is often another profound truth." He continued: "Truth should be a plural. There can be many true statements or ideas or formulae that are nonetheless incommensurable…. I would celebrate multiple incommensurable truths as a sign of human flourishing, not a problem to be solved.”
​
This appreciation for multiple perspectives is a theme that is emerging in research of both biological and artificial intelligence. The more varied the training data, the more rich the diversity of opinions and ideas, the easier it is to discover new truths about the world. By putting our heads together, we can see the big picture a lot easier than if we try to figure it out alone.