Ethan Landes

Ethan Landes

Academic Philosopher

Projects

Innovating Methods for the Ethics of Socially Disruptive Technologies

Does new technology require new approaches to philosophy? Yes, and not least because philosophy itself could use a few methodological innovations. New technologies can strain our existing conceptual repertoir creating, for example, metaphilosophical questions around who, if anyone, has genuine conceptual expertise. The first paper of this project accordingly evaluates the approaches philosophers (and interested psychologists) have used to understand whether trusting AI is conceptually coherent. Not only are these methods lacking, more appropriate methods that we adapted from experimental philosophy of language suggest that philosophers have lost the thread on what trust is to everyone else.

Another project, currently in the late stages of preparation, argues that VR versions of moral dilemmas should be adopted as a method in philosophy due to its ability to help users to reevaluate features of the dilemma.

The Ethics and Efficacy of Label Design

Every idea needs a linguistic vehicle – a serious of sounds and letters that we use to discuss it. What effect do such labels have on our perception of the world and how can we responsibly harness this power when designing language? My experimental work has found that label choice has a subtle but very real effect on perceptions of novel concepts. In other words, what we call something affects how we see something. This put weight to the criticism labels sometimes get for being misleading or harmful. I have argued that this criticism is best understood as a mismatch between what non-linguistic information people are apt to infer from a label and what beliefs would be true, helpful, or otherwise desirable.

These findings have consequences for projects of language activism, especially conceptual engineering. If conceptual engineering aims to improve concepts and language to some political, ethical, or epistemic end, then any intended change can be helped or hindered by the effect of labels. Because labels effects are ultimately individual-specific, conceptual engineering should therefore engage in market research when deciding what label to use.

The Learning from Bullshit Trilogy

Across three papers I explored how we can learn from people or things that bullshit us – that don’t care if what they are telling us is true. The core observation in these works is that even when someone cannot, does not want to, or does not intent to tell us the truth, they can nonetheless show us something is true. The first paper looks at insincerely written works of philosophy, such as philosophical hoaxes, arguing such works are perfectly capable of producing knowledge in readers. This, I argued, has metaphilosophical consequences for what inferences we can draw about the methods of philosophy from works of philosophy (a theme I later returned to in the context of conceptual engineering.)

Then ChatGPT happened. Turning to AI, the next paper turned to the question of artificial moral advisors – AI designed to help people make moral decisions. With collaborators, we argued that even if AI does not in any sense know what it is ‘saying’, AI can be used responsibly for moral growth. However, it is only responsible if users use the AI as inspiration for their own thinking and moral growth rather than simply deferring to its advice (see also this post at the APA blog an account generalized beyond ethics). The final paper in the trilogy experimentally tested this distinction in epistemic stances taken towards artificial moral advice. By manipulating the justification for moral advice and participants’ evidence of the moral advisor’s reliability, we found that people who are persuaded by AI moral advice do so because they do in fact defer to the advice. Therefore, intervention should be taken to help develop norms around epistemically responsible AI use.

The Methodology of Conceptual Engineering

At rock bottom, conceptual engineering is about improving how people talk or think. This project explores how conceptual engineering needs to be an empirical process in order to succeed. One strand argues from the armchair that conceptual engineering requires novel methodology approaches. As a practice, conceptual engineers should adopt entirely new methods than they currently use if they are serious about creating real-world impact. At the same time, if we are to understand conceptual engineering as a historical tradition in philosophy, there is no one method by which we should study it. Another strand of the project develops exactly the sort of empirical frameworks the first strand argues conceptual engineers need. Perhaps the most pressing need is a way to test whether conceptual revision occurs. Otherwise conceptual engineers are left guessing whether their intended changes are successful. The uptake of new concepts is easier to spot, but how do conceptual engineers stack the odds in their favor? As I both argue and experimentally demonstrate, one way is by careful choice of what the concept is called.