After reading Iris Meredith’s There’s no such thing as Just a Tool, I was reminded of some related work.

If you’ve been thinking along Torvalds’ line (it’s “just a tool”), you might be interested in reflecting upon the above article. Meredith gives a fascinating analysis of how human–tool combinations are not value-neutral despite how we might idealize them. I see clear dialogue here with questions like “Can you separate the creator from their work?” and “How do we embody values in our technology? Who gets to choose what values are embodied?”

Related work along those themes:

• Greg Wilson’s “Rainy Day Thoughts on AI” (first mentioned in Digital Etiquette) opens with Orwell & Dali (creator vs. work)
• Ruha Benjamin’s excellent book Race After Technology provokes, prods, and pokes many of these questions with recent examples, including software built on statistical models (from, e.g., predictive policing and other legal system applications to things as supposedly innocuous as Google Maps).
• I naïvely engaged this question (or maybe sidestepped it) when thinking about LLMs in January: I took the “tool conversation” (how to use an LLM effectively) for granted and wished to set it aside for a different but related conversation about power and harm. Thanks to this post’s impetus, I’m now reconsidering that the 2 conversations fit in the same framework—using an LLM engenders different priorities and goals, which are evidently related to power and harm. We exercise power to achieve our goals, and we might want to decide how to do so while minimizing harm.

I also stumbled across a few related articles (least useful to most):

• Bryan Tan wrote AI is not “just a tool”. He asks a few ok questions:

  What does the tool do? What is the scope of its operations? What will it replace? How will it effect our lifestyles?

  But, in my view, he performs a kind of criti-hype, which Cory Doctorow has introduced me to more than once. Tan writes

  moralizing about technology can be beside the point. It isn’t necessary in this discussion to talk about tech’s inherent goodness or badness. We can limit ourselves to pondering its effects. We just have to turn our eyes away from how these tools will enable us to shape the world, and start looking at how using these tools will shape us.

  and then further, disguised in what appears on the surface to be critical, promotes the technology:

  We’re in the nascent days of this technology, where the products that are going to shape the future are just now hitting the public at scale in their most primitive forms. We’re at a critical juncture where we need to think deeply about what some of these effects might be.

  […]

  [AI’s] arrival brings not one new medium to reckon with, but countless, and the obsolescence of just as many.

  So rather than engage with the good questions (such as how we are shaped by the human–tool gestalt), he asserts that “AI” will obsolete many media and create new ones, that it will shape the future. I suppose that’s vague enough to be hard to falsify, and thus probably true. Yet it can’t help but leave the “tool” question short.

• Christine Haskell wrote AI Is Not Just A Tool, which I think wants to make a deep point: most tools engage us with our reality, while “AI” proposes (or invents!) a different reality. That’s a really interesting thesis, from which we ask more questions: who determines the shapes of those realities? Whose values are thus embodied? What might reality distortions through the lens of these tools do to our psyches?

  Haskell goes on to write about legal and regulatory differences between a tool and a system, which naturally calls into question responsibility. Are these things inherently accountability sinks, or are there other modes? What do we do when things go wrong? Better to understand that “systems acting on the public [should be] answerable to the public.” Yet today it so often seems our public systems place vastly more burden on harmed individuals, and make it harder for them to claim and redress harm, than they do to regulate and hold accountable said systems. (Fortunately, this is a matter of policy, and policy can change.)

  But here’s why I still find this engagement with the tool question not as helpful as others: from the second paragraph onward the writing reeks of LLM output. I nearly didn’t finish the article, and many sentences become meaningless for their flowery. I find this disappointing from such an intellectually distinguished individual who prides herself on “keeping values in the room.” It’s not that the LLM can’t be helpful in writing; it’s that the generated text so clearly distracts from or obscures her useful points. I have a hard time discerning which values are in the room with me.

• Finally, HCI and data visualization and accessibility researcher Frank Elavsky wrote Stop saying that AI is just a tool and it only matters how it is used. Watch the whole talk, too.

  Unsurprisingly, this article takes as given that “artifacts have politics” and interrogates how “just a tool” is used. In Elavsky’s conception, it is “deceptively misleading,” “deceptively simple,” and “insultingly naive”: in other words, might it only serve to shut down or limit the conversation, just as Meredith writes? Uttering this phrase doesn’t contribute anything meaningful to our current discourse.

  Elavsky is also concerned by the usual follow up, “it matters how you use it.” Besides refusing to engage with centuries of thinking about morality (as do the answers at time of writing to the StackExchange question “Ethics of using AI in creative work”—no wonder Felienne Hermans is so concerned with CS not engaging with other fields!), this phrase also participates in a tradition of individualism that some credit with destroying community and blinding us to structural, systematic change that might solve some of our most pressing problems.

  I am thus utterly unsurprised that Elavsky also cites Heidegger as Meredith did.

  I won’t summarize all (the rest) of Elavsky’s engagement with this question (sneak peaks: getting rid of all pain and flattening our being are not good results; find the street’s use for these things; “go and do what machines cannot”). Really, you should go read. Engage with the question yourself: what tools shape you, and how?

Categories: Blog

Previous Next

Back to posts