Tessera

Travis LaCroix

@travislacroix.bsky.social

Dr // Asst. Prof // Philosopher @ Durham University (UK) (I am also a human being) Language Origins // AI Ethics // Autism

99 Posts154 Follower 82 Folgt

Travis LaCroix @travislacroix.bsky.social · 2mo ago

1/ Just published in Higher Ed! We argue that debates about whether GenAI use in universities is "academic misconduct" are asking the wrong question : Before deciding what counts as misconduct, universities need to be clear about what they think universities are for. doi.org/10.1007/s107...

doi.org

British Journal for the Philosophy of Science @thebjps.bsky.social · 2mo ago

New from the BJPS Review of Books: Artificial Intelligence and the Value Alignment Problem – Travis LaCroix Reviewed by Rune Nyrup www.thebsps.org/reviewofbook... #philsci #philsky

Travis LaCroix, Artificial Intelligence and the Value Alignment Problem | BJPS Review of Books

Rune Nyrup reviews Artificial Intelligence and the Value Alignment Problem, by Travis LaCroix

thebsps.org

Travis LaCroix @travislacroix.bsky.social · 3mo ago

Sad that this paper came out after we wrote ours on strategic polysemy, because "emotion vector" and "functional emotions" are great case studies in *glosslighting*.

Although we have focused on just a few instances of the same phenomenon here, many other terms have similar characteristics as those described above—e.g., consciousness, foundation model, general intelligence, beliefs / preferences / goals, creativity, planning / deliberation, goal-directed behaviour, empathy / affective AI, trustworthiness / honesty, personality, scheming, etc. (See Appendix A.) Across these examples, the same pattern recurs, highlighting how these terms serve a dual purpose. They signal one thing to the public, policymakers, and investors—typically the intuitive, anthropomorphic, or metaphorical meaning—while simultaneously retaining a narrow, technical interpretation to which experts can retreat when challenged. This systematic ambiguity enables, what we call, glosslighting.

Glosslight (Definition):
Glosslighting (verb) is the practice of using technically redefined or polysemous terms to evoke familiar meanings—often emotionally or cognitively powerful ones—while preserving the ability to deny those meanings through retreat into specialised, context-bound reinterpretations.

The rhetorical effects of glosslighting—suggesting familiar meanings while retaining deniability—may arise intentionally, but they can also emerge from (i.e., are a systemic outcome of) the foreseeable interaction between ambiguous terminology (allowing multiple interpretations), and heterogeneous audiences (different groups interpreting the same term differently), combined with institutional incentives (encouraging strategic ambiguity).

Emily M. Bender@emilymbender.bsky.social · 3mo ago

Reading Anthropic's recent nonsense about "emotion vectors" and was struck by this remark. Is there really such a taboo? Because we see anthropomorphizing language about "AI" systems *all the time*.

Screencap:

The case for taking anthropomorphic reasoning seriously
There is a well-established taboo against anthropomorphizing AI systems. This caution is often warranted: attributing human emotions to language models can lead to misplaced trust or over-attachment. But our findings suggest that there may also be risks from failing to apply some degree of anthropomorphic reasoning to models. As discussed above, when users interact with AI models, they are typically interacting with a character (Claude in our case) being played by the model, whose characteristics are derived from human archetypes. From this perspective, it is natural for models to have developed internal machinery to emulate human-like psychological characteristics, and for the character they play to make use of this machinery. To understand these models’ behavior, anthropomorphic reasoning is essential.

Source: https://www.anthropic.com/research/emotion-concepts-function

Sentence starting with "There is a" highlighted in red.

Travis LaCroix @travislacroix.bsky.social · 3mo ago

New paper pre-print, accepted at FAccT 2026, wherein @fintanmallory.com , @sashamtl.bsky.social , and I argue that hype-laden AI terms like “hallucination”, “reasoning”, and “agent” are doing strategic, rhetorical work, which we call "glosslighting".

Image of title, authors, and abstract for "Strategic Polysemy in AI Discourse: A Philosophical Analysis of Language, Hype, and Power" (Travis LaCroix, Fintan Mallory, and Sasha Luccioni).

"This paper examines the strategic use of language in contemporary artificial intelligence (AI) discourse, focusing on the wide- spread adoption of metaphorical or colloquial terms like “hallucination”, “chain-of-thought”, “introspection”, “language model”, “alignment”, and “agent”. We argue that many such terms exhibit strategic polysemy: they sustain multiple interpretations simultaneously, combining narrow technical definitions with broader anthropomorphic or common-sense associations. In contemporary AI research and deployment contexts, this semantic flexibility produces significant institutional and discursive effects, shaping how AI systems are understood by researchers, policymakers, funders, and the public. To analyse this phe- nomenon, we introduce the concept of glosslighting: the practice of using technically redefined terms to evoke intuitive—often anthropomorphic or misleading—associations while preserving plausible deniability through restricted technical definitions. Glosslighting enables actors to benefit from the persuasive force of familiar language while maintaining the ability to retreat to narrower definitions when challenged. We argue that this practice contributes to AI hype cycles, facilitates the mobilisation of investment and institutional support, and influences public and policy perceptions of AI systems, while often deflecting epistemic and ethical scrutiny. By examining the linguistic dynamics of glosslighting and strategic polysemy, the paper highlights how language itself functions as a sociotechnical mechanism shaping the development and governance of AI."

Travis LaCroix @travislacroix.bsky.social · 5mo ago

altmetric sentiment analysis is so funny. It's always like "49% of mentions of your work are STRONG NEGATIVE", and then it's just my own posts on BlueSky.

Travis LaCroix @travislacroix.bsky.social · 5mo ago

Just published in Psychological Inquiry! I offer a sustained philosophical and empirical critique of the theory-of-mind-deficit explanation of autism. The ultimate conclusion is that the research programme has become degenerative—and therefore pseudoscientific. www.tandfonline.com/doi/full/10....

Autism and the Pseudoscience of Mind

The theory-of-mind-deficit explanation of autism proposes that autistics lack a theory of mind, that autism comprises a theory-of-mind deficit (strong version); or, that autistics often have diffic...

tandfonline.com

Travis LaCroix @travislacroix.bsky.social · last yr.

What do philosophers talk about when they talk about autism? Usually: not actually autism. When they do talk about autism, they often spread harmful stereotypes. Our scoping literature review, now published (open access) in Synthese, analyses the “philosophy of autism” subfield. rdcu.be/ewIAr

What do philosophers talk about when they talk about autism?

rdcu.be

Travis LaCroix @travislacroix.bsky.social · last yr.

I didn't announce this when it happened, but my article, "Metaethical perspectives for 'Benchmarking' AI ethics", with @sashamtl.bsky.social, was published back in March in the AI and Ethics journal. doi.org/10.1007/s436...

Metaethical perspectives on ‘benchmarking’ AI ethics - AI and Ethics

Benchmarks are seen as the cornerstone for measuring technical progress in artificial intelligence (AI) research and have been developed for a variety of tasks ranging from question answering to emoti...

doi.org

Travis LaCroix @travislacroix.bsky.social · last yr.

When people suggest that maybe AI can feel emotions, these are the emotions they are talking about.

cait (and adonis)@cait.bsky.social · last yr.

a friend of mine shared this ai-generated "emotion wheel" and unfortunately i have been laughing my ass off at it for like 15 minutes now. today i am feeling Fnliinneon

emotion wheel starting with "joy, fear, anger" in the center and then lapsing into gibberish

Travis LaCroix @travislacroix.bsky.social · last yr.

Currently diving deep into the recently-tabled Bill C-2. There is a lot going on here. The bill would allow officials to cancel, suspend or change immigration documents in the 'public interest'. Critics have quickly voiced worries about asylum-seekers and refugees.

Travis LaCroix @travislacroix.bsky.social · last yr.

I am disappointed to report that the 2018 film, The Predator, does not treat neurodiversity or disability with any sensitivity or nuance.

Travis LaCroix @travislacroix.bsky.social · last yr.

Anyway, all this forms the foundation for my book-length analysis of how to properly conceive of value alignment and misalignment so that these words are meaningful and useful. broadviewpress.com/product/arti...

Artificial Intelligence and the Value Alignment Problem - Broadview Press

Artificial Intelligence and the Value Alignment Problem -

broadviewpress.com

Travis LaCroix@travislacroix.bsky.social · last yr.

To try to fix the hot mess that is the field of value alignment, I give a new description of the problem based on the principal-agent problem from economics. The value alignment problem is a class of problems, which is instantiated when we delegate tasks to AI systems.

Screen shot of a page of a book ("Artificial Intelligence and the Value Alignment Problem"), which reads:

"The Value Alignment Problem (Structural Definition)
A problem that arises from the dynamics of multi-agent interactions involving the delegation of tasks from one actor (a human principal) to another (an AI agent). This problem can arise whenever

(a) The agent’s objective function is misaligned with the true objective of the principal(s); or,
(b) There are informational asymmetries between the principal and the agent."

Travis LaCroix @travislacroix.bsky.social · last yr.

I agree with @abeba.bsky.social here; but, for better or worse, I am trying to fix the hot mess that is the field of value alignment! 😬

Screenshot of a page of a book ("Artificial Intelligence and the Value Alignment Problem"), which reads:

"Thus, although there is no authoritative definition of the value alignment problem, a somewhat standard specification can be formulated as follows:

The Value Alignment Problem (Standard Definition)
The problem of ensuring that AI systems are aligned with the values of humanity.

Whether explicitly stated or merely implied, this intuitive specification of the value alignment problem for artificial intelligence is relatively common in discussions on the topic. Moreover, the standard definition is superficially useful as an initial description of the problem. On the surface level, it seems relatively simple to understand. However, closer examination or analysis makes apparent that this conceptualisation underspecifies the problem; hence, this definition is of no conceptual use for practical work on mitigating value misalignment."

Dr Abeba Birhane@abeba.blacksky.app · last yr.

the entire value alignment field is a hot mess

Travis LaCroix @travislacroix.bsky.social · last yr.

As I am working on final (and so finicky) edits (including spacing and typesetting) for my book, I maintain that LaTeX's slogan should be: "It makes the difficult things easy, and the the easy things difficult".

Travis LaCroix @travislacroix.bsky.social · last yr.

Classic from the archives.

A three-panel meme featuring a scene from Donnie Darko.

First panel: Donnie (played by Jake Gyllenhaal, and labelled in the image as "evolution") looks serious. Text [evolution]: "I made a new animal".

Second panel: An older woman (his therapist) sits in a chair, looking skeptical. Text: "new animal or crab again?"

Third panel: "crab again."

Travis LaCroix @travislacroix.bsky.social · last yr.

Give yourself a present? Once a day? EVERY day?! In THIS economy?!?

Travis LaCroix @travislacroix.bsky.social · last yr.

Me: "How many times shall I forgive my [department] who sins against me (asking me to do extra grading)? Up to seven times?" My Department: "I tell you, not seven times, but seventy times seven."

Travis LaCroix @travislacroix.bsky.social · 2y ago

This is a very good book, and a fun read! Continues the excellent work that @emilymbender.bsky.social and @alexhanna.bsky.social have been up to. Definitely get a copy!

Alex Hanna@alexhanna.bsky.social · 2y ago

Pre-order alert, cover, and website reveal for our book, THE AI CON! (with @emilymbender.bsky.social) It's out for release on May 13, 2025! thecon.ai

Travis LaCroix @travislacroix.bsky.social · 2y ago

This article has (finally) been published (open access!) in Philosophical Studies! The Linguistic DEAD ZONE of Value Aligned Agency, Natural and Artificial. doi.org/10.1007/s110...

The linguistic dead zone of value-aligned agency, natural and artificial - Philosophical Studies

The value alignment problem for artificial intelligence (AI) asks how we can ensure that the “values”—i.e., objective functions—of artificial systems are aligned with the values of humanity. In this p...

doi.org

Travis LaCroix @travislacroix.bsky.social · 2y ago

Me, once again, departing Calgary

a stack of orange boxes are sitting on a carpet in a room

Alt: Reversed gif of Woody from Toy Story 2 entering a room from a box, while tipping his hat

media.tenor.com

Travis LaCroix @travislacroix.bsky.social · 2y ago

Me returning to Calgary:

an orange box is sitting on a carpet in a living room

Alt: Gif of Woody from Toy Story 2 entering a room from a box, while tipping his hat

media.tenor.com

Travis LaCroix @travislacroix.bsky.social · 2y ago

You ever think about how 0.014% of Charlottetown* is in the band Alvvays? (*Circa 2011)