reclaim yourself from recommendation algorithms

fighting algorithms with other algorithms (it makes sense, I promise)

I am increasingly using AI as a weapon against all the algorithms on the internet that are trying to separate me from my attention, my money, and any sense of an identity that is not recommended to me. I find this interesting! Fighting algorithms with algorithms. Who am I.

Well, let me back up.

Over the past 15 years, algorithms (and particularly recommendation algorithms) have become invisible architects of our desires. Beyond predicting what we want, they actively shape it. Many smart critics have written about how this has transformed our world into a bland sameness. We all want to travel to the same “hidden gem” destinations to visit cafes that all look the same and play the same music. Your friends follow an influencer who tags where she bought her clothes so they visit those websites and then YOU get targeted on those websites and suddenly this random shirt is following you around the internet. Etc.

I don’t think this is something inherently bad about, or unique to, algorithms. Longtime readers of this newsletter may remember my essay about Mad Men, where I opined on the yin and yang of Don Draper, a man clinically incapable of having, only of wanting, whose only success in his life came from his ability to manufacture desire in others at scale. Advertising has been around for ages! And even advertising is not inherently bad.

But. The house of mirrors fools us all. Do we want a thing, or have we just been convinced we want it? Do I even like mid century furniture? Was I always low key obsessed with loafers? Is it a coincidence that they are everywhere now? I swear I was wearing them before they were everywhere. But honestly I have no idea.

What I do know is that none of these systems are designed to help you be more authentically you. They are optimizing for two things: extract your attention or extract your money. You don’t get to choose what they optimize for. But with AI, to an extent, you do get to choose. You can create a stunningly intelligent tool that exists to serve whatever objective you want. Instead of being the subject of algorithms, you can be their architect.

You can create GPTs that help you:

1. Come up with reasons not to purchase something

2. Filter potential purchases through your unique style preferences to help you de-influence yourself

3. Evaluate vacation destinations based on the specific experiences you value

4. Develop the vocabulary to talk about what interior design concepts you actually like, not just what’s trending on Pinterest

5. Discover music based on your esoteric tastes rather than what’s popular with “people like you”

To build tools that successfully do this, there are a few things you need to keep in mind:

1. Start with your likes/dislikes: Instead of asking ChatGPT "What do you recommend?" , start with "Here's what I actually like and why." This way, your preferences provide the foundation for everything that follows.

2. Clearly define what you are optimizing for: Commercial recommendation algorithms optimize for engagement and conversion. You can make a Custom GPT that optimizes for anything: genuine satisfaction, alignment with your values, long-term utility, deeper understanding, or whatever matters to you.

3. Get creative with different outputs: Do you need gentle pushback against impulsive decisions? Structured analysis of your preferences? Probing questions that help clarify your thinking? All of these are possible.

This might all feel abstract. Allow me to make it concrete. Let’s say you tend to buy clothes when you’re out at a store but when you get home you realize they don’t really fit with the rest of your wardrobe. I get it, I too have suffered delusions of rebuilding my entire personality due to good mirrors, flattering lighting, and a charming salesperson. So, you want to make a Custom GPT that will compellingly talk you out of buying things.

Here’s how I would go about this:

1. Feed it data about your specific tastes. LIterally, upload a bunch of photos of outfits you always gravitate toward, items in your closet, your measurements. If you have the vocabulary to explain what you like, do that, but you can also just feed it the raw data and then say, “based on what I’ve told you, describe my style in extremely specific detail, including shopping mistakes I am likely to make.”

2. Clarify the objective: “I want to create a custom GPT that will persuasively attempt to talk me out of buying anything that is not an obvious must-buy. What approach do you recommend I take?”

3. Read its suggestion and offer feedback. “I don’t want it to have a fun and cheeky tone. That will annoy me. I want it to be extremely persuasive in its arguments without resorting to rhetorical tricks.”

4. Ask it to write a prompt for you. Load up that prompt, and your training materials from step 1, into a Custom GPT or Project.

5. Test it by uploading a photo of something you are considering buying. See what response you get. Go back to your chat where it gave you the prompt. Tell it what you like or don’t like about the response you got and ask it to rewrite the prompt (or do this yourself).

6. Voila! You have created a turncoat algorithm who will fight alongside you in the Great Recommendation Algorithm Wars.

I understand that we as consumers do not actually have “full” control over how these AI models behave. But these tools are incredibly adept at finding connections and patterns in data. Most people I know struggle to see the patterns in their choices, preferences, and behaviors.

This is the part where we've been vulnerable to manipulation. We can't see our own patterns, but the algorithms tracking us absolutely can. They know what makes us click, what makes us linger, what makes us buy. They've been using pattern recognition against us for years.

So why not flip the script? Use pattern recognition to actually understand yourself better, and build tools that help you refine and defend your authentic tastes. Maybe you’ll learn something interesting along the way.

Miscellaneous Recommendations

1. Survivor, the best show on TV, kicked off its 48th season last week. A few years ago I wrote a letter of recommendation for why you should be watching Survivor. It is slightly out of date now but still gets the job done. This season is shaping up to be a good one. You can watch it on Paramount+.

2. I am continuing to work my way through Justine Doiron’s cookbook. On Sunday I made the Spoon Salad (more of a grain bowl than a salad) and it was one of the best things I’ve eaten all year. I looked for the recipe online and found it but you have to create a today.com account to see it, lol. Maybe just get the cookbook.

xoxo,

hils