Algorithmic Comparison

A simple explanation

You did not choose the comparison-set. A recommender chose it for you, optimising for what would keep you scrolling rather than for what would be useful to know. The Belonging System, asked whether your location in the group was safe, was handed a fabricated group: strangers you have never met, peers you barely know, public figures whose lives are visible only at their peaks, all assembled because their content produces engagement in people like you. The System cannot tell the difference between this cohort and a real one. It runs the same machinery against the feed that it would run against your village.

This is what makes algorithmic comparison structurally different from older comparison loops. The mechanism is identical — felt-sense of relative position, residue accumulation, foreclosed location-question — but the throughput is orders of magnitude larger, and the comparison-objects were selected against you rather than encountered by you.

An everyday example

You open the app at 9:12 a.m. for one specific reason — to check a message. In the eleven seconds before you find the message, the feed has presented six comparison-objects: a milestone post, a body, a holiday, a project launch, a casual flex about ease, a thread about someone's success. Each one was selected for the moment by a model trained on what kept you specifically engaged in the past. Your Belonging System has registered six felt-senses of relative position before your conscious attention has even located the message.

By 9:13 a.m. you have closed the app. You feel a small, diffuse low-grade flatness you cannot trace. You spend the next twenty minutes trying to do the thing you actually intended to do that morning. The work feels smaller than it did before 9:12. By evening you have repeated the sequence six or seven times, and the day's residue is a tonic flatness that has no nameable cause because there were too many causes to name.

Why does the algorithm keep showing me people I find triggering?

Because the recommender's objective function is engagement, not your wellbeing, and high-contrast comparison-objects are some of the most reliably engaging content for the Belonging System. You watched the post slightly longer last time. You clicked through. You looked at a comment. The model logged the engagement and gave you more of what produced it. The fact that the engagement was driven by Belonging anxiety rather than by enjoyment is invisible to the system. They look identical from the outside.

This is what makes algorithmic comparison a category of its own. In older comparison loops, you encountered your peer-set by walking through the world; the comparison was a side effect of being in a group. In algorithmic comparison, the comparison is the product. The feed has been optimised to deliver as many comparison-triggers per minute as can fit, because comparison-triggers are reliably engaging. The mechanism inside you is not new. The dosing is.

The behavioral loop

A loop that hides because each scroll feels brief:

1. Trigger — the app opens, or the feed refreshes; within seconds, the recommender presents a high-contrast comparison-object.
2. Scan and select — the System reads the comparison-object before conscious attention has caught up.
3. Belonging verdict — a felt-sense of relative position is issued.
4. Substitute feeling — a faint inadequacy, envy, or low-grade flatness arrives.
5. Continued scrolling — the same surface delivers the next comparison-object within seconds; the loop runs again.
6. Brief clarity — eventually the session ends, often without a memory of any individual comparison-object.
7. Residue — tonic flatness, diffuse self-distrust, an attention that has been narrowed and is slow to widen back; the residue cannot be traced to any single trigger because the triggers were too numerous.
8. Re-entry — the next opening of the app produces faster engagement, because the body has been conditioned to expect the dose.

Emotional drivers

Four feelings, often stacked:

• A diffuse low-grade flatness that cannot be traced to a specific comparison, because the throughput was too high for any single comparison to be visible.
• A faint compulsion to keep scrolling, often felt as not-quite-boredom, which is the body seeking the next dose.
• A growing irritation at the feed itself, often re-coded as critique of the platform rather than recognised as residue.
• An anticipatory wariness about opening the app, which is overridden by the same opening behaviour within seconds.

What your nervous system does

The trigger of a high-throughput feed produces a sustained mid-grade sympathetic activation — narrowed attention, shallow breath, mild forward lean, low-grade glucose mobilisation. The body reads the feed as a high-stimulus social environment and prepares accordingly. The activation is too low to be felt as stress and too constant to register as a discrete event. It is just what scrolling feels like.

Over months and years, the activation becomes a baseline. The body's resting attentional posture narrows. Recovery times lengthen — the slow widening of attention after a session takes longer than it used to. People in this state report that their unscrolled hours feel less interesting than they used to, not because the hours changed but because the baseline of stimulation has been re-set.

The DojoWell interpretation

Algorithmic comparison is the textbook effort_without_deposit density signature. The effort, measured in hours per week of scanning, processing, and recovering, is enormous. The deposit is near-zero, because the comparison-objects were selected for engagement rather than for any quality that would integrate into your life. The System is asked to perform location-readings against a cohort that has nothing to do with your actual cohort. The verdicts it produces cannot inform a decision, a skill, or a relational move, because the comparison-objects are not in your social field in any actionable sense.

The substitution mechanism is the same as in upward, lateral, or downward comparison — a felt-sense of relative position standing in for the Belonging System's original location-question. What is different is that the loop has been industrialised. A pre-modern System was asked to read perhaps a dozen social signals a day. A modern System on a default feed configuration is asked to read several thousand. The mechanism cannot scale that far. The result is not better social information; it is residue at industrial throughput.

This is also why algorithmic comparison is the most efficient way to lower density at scale in modern life. The throughput compounds across days, weeks, and years. The unmet location-question never reaches the relational field because the relational bandwidth has been spent on strangers selected by a model. The body knows something is wrong but cannot point at any single comparison-object as the cause, because the cause was the volume.

The work is not to feel better about each comparison. It is to recognise that the loop's signature is throughput, and that the cheapest intervention is at the source.

How do I detox from algorithmic comparison?

You do not need to argue with the comparisons. You change the dosage. The System will continue to issue verdicts on whatever it is shown; what is workable is how much it is shown.

Three moves, in order of difficulty:

1. Cut the throughput. Halving the daily session time, or halving the number of feed-surfaces in your day, reliably reduces the loop. This is the cheapest intervention by a large margin.
2. Pre-decide the entry. Open the app for the thing you came for and close it before the feed has time to dose you. The pre-decision is what survives the friction of being mid-scroll.
3. Track the recovery, not the comparisons. A week of measured post-session flatness is data the loop-runner can use; an attempt to count comparisons is futile.

Practical steps

1. Audit the throughput. Most people underestimate their feed time by half. A week of honest measurement reveals what the loop is actually being fed.
2. Reduce feed-surfaces. Each surface compounds; halving the number of apps that can dose you is more effective than rationing each one.
3. Move comparison-heavy apps off the home screen. A two-second friction at the entry point cuts a meaningful fraction of opens.
4. Install one analogue hour per day. A walk, a meal, a conversation without the phone in reach. The unscrolled hour resets the baseline.
5. Track the residue across a week. Sleep quality, post-session attention recovery, evening mood. The residue is data; the comparisons are not.

Reflection questions

• What time of day does the feed dose you most heavily, and what does the residue cost the rest of that day?
• How do I tell the difference between curiosity and the body seeking the next comparison dose?
• Whose comparison-objects does the algorithm most reliably surface to you, and what does the set suggest about your underlying location-question?
• Where would you put the hours you currently spend in algorithmically dosed comparison if you got them back?

Frequently Asked Questions