LESSWRONG
LW

1106
Jordan Rubin
42840
Message
Dialogue
Subscribe

Researcher-Operator currently on garden leave. Formerly: Two Sigma (Quant Research + Mgmt) / OnDeck (Data science in lending) / BlackRock (Bond desk quant). I hope my thinking can be helpful to you!

My Substack: https://jordanmrubin.substack.com

My LinkedIn: https://www.linkedin.com/in/jordanmrubin/

Posts

Sorted by New

Wikitag Contributions

Comments

Sorted by
Newest
No wikitag contributions to display.
Follow-up to "My Empathy Is Rarely Kind"
Jordan Rubin2mo30

Empathy is a multi-dimensional operation and there are many ways to do it. Here are some examples of dimensions along which you can vary empathy:

  • How much do you try to literally “feel their feelings” vs feel your own feelings vs feel no feelings?
  • How charitably (or hostilely) do you try to read their situation?
  • How focused are you on actions they could take?
  • To what extent do you try to understand the story of what took them here vs focus on the present moment?
  • How focused are you on what is good for them vs on what is good for both?

    You don’t have to do empathy the same way in every situation. But have you ever tried to max out on “feel their feelings” or “charitable read”?

    Here is o3’s charitable read of nail-woman:
     
  • felt-sense override – pain is embodied confusion. before she can orient to “extract object,” she needs another nervous system to mirror how disorienting the signal is. otherwise her interoceptive map keeps screaming “unsafe,” so any fix attempt feels like an attack on the only reliable data she has left.
  • status & competence threat – accepting a drive-by solution telegraphs “i couldn’t notice a metal spike in my own skull.” that’s ego-annihilating. she’s fighting for narrative sovereignty, not against physics.
  • pattern-matching past dismissals – likely history of “just calm down” responses. the nail scene retriggers that cached pattern; she’s guarding against another instance of emotional rug-pull.
  • speech-act mismatch – she’s broadcasting affective data; partner keeps replying with instrumental commands. protocol violation feels like deafness. she insists “it’s not about the nail” bc rn the channel is meta-communication, not mechanics.
Reply
Dimensionalization
Jordan Rubin3mo00

Let's do one example. Suppose you are trying to get access to celebrity participants for your music videos, since we've identified that as high leverage. Let's start by listing some things that could meaningfully lead to celebrities agreeing to participate:

  • One of the celebrity's friends recommends that they work with you
  • The celebrity sees your video and wants their work treated similarly
  • Your agent connects with the celebrity's agent and arranges a quid pro quo
  • You use archive footage of a celebrity in your video and get their permission after the fact
  • ...

Now we can filter for what is in your control:

  • You can try to befriend a celebrity's friends
  • You can try cold-sending your work to celebrities
  • You can hire an agent and ask that they help you work with celebrities
  • You can try to find footage of celebrities who might be amenable to this kind of fair use
  • ...

Of these, hiring an agent strikes me as the most certain way to achieve the outcome.

If we had like 50 options (we can get that by using an LLM), then it would make sense to group them and optimize for Fidelity and Complexity. But for now we are focused solely on a few high-leverage actions and I'm not using an LLM, so I won't go through the whole dimensionalization process.

If you again find yourself thinking "but I can't befriend a celebrity's friends" or something like that, you can just do the exact same exercise to dimensionalize that outcome.

Reply
Dimensionalization
Jordan Rubin3mo10

Thanks. Here's what I would suggest if you want to do Dimensionalization by hand starting with Leverage:

  1. Start with actions you can take, and filter for those that would result in things that matter to your outcome. Or,
  2. Work backwards from things that would matter to your outcome, and filter for those that have triggers that are at least partially in your control.

Then, once you've made a list of atomic actions with high Leverage, you can group them in various ways to reduce Complexity and align to your model of 'how the outcome works' to increase Fidelity.

FWIW, it feels obvious to me that having celebrities participate in your videos lends credibility (strong social proof) that could lead to additional commissions. Fine if your target commission audience would be put off by this; in that case, you should find audience-relevant signals of credibility. But credibility is probably going to be important regardless of how you achieve it.

Reply
Dimensionalization
Jordan Rubin3mo10

Hi -- I love the question. 

First, it matters what you choose to dimensionalize: "Dimensionalize Posting Instagram Content" will give you different results from "Dimensionalize Soliciting More Commissions". And "Dimensionalize Posting Instagram Content with respect to its impact on Soliciting More Commissions" will give you still different results. The key is that dimensionalization tries to give you levers to affect a specific decision or outcome.

Second, I strongly recommend using LLMs to help you dimensionalize. Without domain expertise in these areas, I am not going to provide much edge vs an LLM. You can take your comment, paste it into an LLM, link to this post, and ask for dimensionalization of whatever you want using this framework. Part of the goal of the post is to make it easy for me (and others) to do this :)

Here's an example, straight from o3:

axis0 ←—— dial ——→ 5why it matters
authority signalpersonal art dump → visibly paid, tagged celebs, bts w/ shotlist overlaycommissions need trust; ig ranks “creator credibility” too 
genre fitrandom aesthetic soup → laser‑targeted to niches you wantclients self‑identify; boosts save+share rate, a heavy reel weight 
commissionability clarityno hint you’re for hire → explicit cta (“dm ‘video’”) + pricing highlightconversion > likes; CTAs crank follower→client funnel 
hook strengthslow fade‑in → 0‑2 s punch, clear conceptcompletion rate & replays dominate reel ranking 
trend leverageobscure track → trending audio + remix stickersurfing trend boosts discoverability; recency is a top signal 
social proof densityno comments → artist shout‑outs, collab tags, fan duetsexternal validation both ranks & persuades
format hygiene16:9, long captions → 9:16, 6‑15 s, native edits, closed‑capsalgorithm favors “made‑for‑reel” metadata 
network activationpassive post → “tag the band you’d cast here” promptcomments + shares multiply reach weight 
portfolio breadthsingle style → carousel of radically different looksshows adaptability, dissolves “can they do MY vibe?” objection

You can then ask for candidate videos (or provide your own) and have the LLM rank by dimension:

ideaauthoritygenre_fitcommissionabilityhook_strengthtrend_leveragesocial_proofformat_hygienenetwork_activationportfolio_breadth
30s reel: behind‑the‑scenes glam rock shoot w/ artist tag + shotlist overlay443324432
12s reel: fast‑cut genre demo montage (pop, indie, hiphop) w/ trending sped‑up audio253553445
15s testimonial: happy client artist talking on‑cam, intercut finished shots, clear 'dm to book'545425432
8s hook: extreme macro sparkle shot ending w/ logo reveal, trending glam track342542421
carousel: storyboard frames + final shots for recent commission, caption explains process434313323
reel: split‑screen 'before vs after' color grade on mobile footage vs final cut334423423
story highlight: pricing tiers infographic + link sticker to inquiry form435213321
15s concept pitch: 'which band fits this visual?' prompt, encourages tagging244434353
9s reel: top 3 camera angle tricks demoed w/ on‑screen text333432424
12s reel: collab remix with popular cinematographer duet, shows adaptability443445444

If you don't like the dimensions or the examples, you can ask for new ones, or pick a different choice to dimensionalize.

Hope this is helpful!

Reply
2Conceptual Rhyme and Metaphor
1mo
0
8Sharpening the Shears: 8 Lessons from Garden Leave
2mo
0
3Mapping Mental Moves
2mo
0
10Use AI to Dimensionalize
3mo
1
11Proverbial Corollaries
3mo
0
7Memories of the Neutral Zone
3mo
0
5Dimensionalizing Forecast Value
3mo
0
7Dimensionalization
4mo
6