The Wrong Attractor · Modern Psychology

Why is quitting always so much harder than starting?

Anyone who's tried to leave a habit, a substance, or a loop knows the asymmetry in their body: getting in was quick and easy; getting out is long and brutal. That gap isn't weakness — it's the whole subject of this module. Once you see it, addiction and OCD stop looking like failures of willpower and start looking like the same piece of geometry.

Module 1 was about being pushed away from your ground state and the pull back toward home. This one is about what happens when a mind stops coming back — when it settles into a stable place that isn't home and begins defending it as home. Call that a wrong attractor: a false ground state with its own gravity.

Addiction installs one. The first high isn't the trap — that's just a big displacement. The trap is what repetition does underneath: each cycle drags your baseline down, until the level that once felt normal now feels like withdrawal. Your system quietly re-files the drug as home. That's why getting clean feels like the worst displacement of all — leaving a false home is, by definition, the farthest thing from it. And it's why relapse drops you back at your old dose, not a beginner's: the old basin never fully fills in.

OCD is the same shape with a different face. The obsession is the spike; the compulsion looks like coming home — it brings relief — but it's simulated return: it quiets the alarm without switching it off. Worse, completing the ritual quietly confirms the threat was real, so the alarm returns louder. The loop tightens itself. That's why "just stop" never works — each "return" is the very thing deepening the trap.

The clinical punchline is gentle, not condemning: you don't escape a wrong attractor by white-knuckling against it. You escape by reshaping the landscape — draining the false home's pull and building a real ground state strong enough to out-pull it.

What you'll be able to do

Define a wrong attractor $W$ as a stable configuration with $W \neq S^0$, and distinguish it from a transient displacement excursion.
Explain how addiction and OCD install and deepen a wrong attractor, and how this deepening makes the return path to $S^0$ progressively more costly (DC5 irreversibility, broken return path).
Distinguish genuine return to $S^0$ from simulated return, and apply the distinction to relapse and to the self-amplifying obsession–compulsion loop.

The precise version

Everything above, stated formally — wrong attractors, basin depth, and the dynamics of tolerance, withdrawal, and relapse. Go as deep as you like, or move on to the next module.

The displacement framework fixes a ground state $S^0$, a displacement $\xi = \rho(s, S^0)$, and a non-negative cost $D(\xi) \ge 0$ that accumulates as $\Phi = \int_0^T D\,dt$. Module 1 treated displacement from $S^0$. Module 2 treats what happens when a system stops trying to come back — when it settles into a stable configuration that is not $S^0$. That configuration is a wrong attractor $W$: a stable fixed point with $W \neq S^0$ and $D(W) > 0$, whose basin captures the current state so that spontaneous return to $S^0$ requires activation energy exceeding the basin depth $\Delta V = V(S_{\text{saddle}}) - V(W)$.

The addiction papers (paper 193 in the series) develop this for the reward system. The pre-addiction ground state $S^0$ is a homeostatic dopaminergic configuration: stable tonic tone in the nucleus accumbens, normal D2 receptor density, a hedonic setpoint $h_0$ at which natural rewards register, and intact prefrontal control. A substance produces supraphysiological dopamine, so first use is a large positive displacement above $h_0$ — the "high." Crucially, first use is not the wrong attractor; it is acute displacement. The wrong attractor emerges only through repetition. Each use cycle triggers allostatic counteradaptation — D2 downregulation, opponent-process recruitment (CRF, dynorphin), a downward migration of the setpoint $h(t) < h_0$ — and these adaptations do not restore $S^0$. They install a new stable configuration $W_{\text{sub}}$. The setpoint trajectory $h_0 \to h_1 \to \cdots \to h_n$, with $h_{k+1} < h_k$, converges on the attractor setpoint $h_{\mathcal{A}}$. Tolerance is this basin deepening: $d_{\text{eff}}(n) \propto e^{\lambda \Delta V_n}$, so the dose needed for a fixed effect grows exponentially with basin depth.

This is the false ground state: the system now treats $W_{\text{sub}}$ as its reference. Withdrawal is then failed return — the path toward $S^0$ becomes the path of maximum displacement, because the opponent processes calibrated to hold $W_{\text{sub}}$ now operate unopposed and drive the system into a trough below $h_{\mathcal{A}}$. By DC5, the return cost exceeds the entry cost: $\Phi_{\text{return}} > \Phi_{\text{departure}}$. Entry was brief and pleasurable; exit is prolonged and costly, because allostatic reorganization is slower to reverse than to install. This is the broken return path.

Craving is the inter-attractor gradient $\nabla\Phi$ pointing toward $W_{\text{sub}}$ — wanting (incentive salience) felt even when liking is gone. Relapse is the dynamical event in which a recovering state $\xi(t)$ crosses the separatrix back into $\mathcal{B}(W_{\text{sub}})$, after which attractor dynamics pull it home. Because abstinence does not erase $W_{\text{sub}}$ — $\Delta$FosB degrades slowly, dendritic changes persist — the recovered landscape retains hysteresis: $\Delta V^{\text{recovered}}_{\text{entry}} < \Delta V^{\text{naive}}_{\text{entry}}$. Re-entry costs less than first entry (kindling, sensitization), so relapse returns the person to the addict's use pattern, not the naive user's. Recovery is therefore not abstinence alone but the construction of a new $S^0_{\text{rec}}$ whose basin out-competes $W_{\text{sub}}$.

OCD is the same wrong-attractor class with a different signature. Its ground state $S^0_{\text{safe}}$ has proportionate threat appraisal: the estimated danger $\hat{p}$ tracks the true danger $p$. An obsession is a wrong-attractor entry event — a trigger generating $D(\xi_{\text{obs}}) \gg v(\xi_{\text{obs}})$, a displacement spike far in excess of real threat. A compulsion looks like return but is simulated return: it lowers $D(\xi)$ transiently without extinguishing $\hat{p}$. Worse, completing the compulsion confirms the threat was real, so $\hat{p}_{t+1}(\xi) = \hat{p}_t(\xi) + \alpha\,\hat{p}_t(\xi)\,\mathbb{1}[\text{compulsion completed}]$ — a positive feedback that blocks the natural correction $\hat{p}_{t+1} \leftarrow \hat{p}_t - \beta[\hat{p}_t - p(\xi)]$. This makes OCD self-amplifying: the activation threshold falls ($\theta_{n+1} < \theta_n$) while the cost of non-completion rises ($C_{n+1} > C_n$), and the basin deepens with every cycle. The neural substrate is the orbitofrontal–striato–thalamic loop: a tonic "something is wrong" error signal that defective striatal gating fails to suppress and the thalamus re-enters — a standing wave the circuit cannot shut down from within.

Worked example

Maya is three months sober from opioids. Her landscape no longer matches the naive one: years of use deepened $W_{\text{sub}}$ and dragged her setpoint to $h_{\mathcal{A}} < h_0$. In protracted abstinence she is on the return path but has not arrived — D2 density and tonic tone normalize only over months to years — so ordinary rewards feel flat (anhedonia: $h(t) < h_0$). One evening she passes the corner where she used to buy. The cue acts as a local landscape deformer, steepening $\nabla\Phi$ toward $W_{\text{sub}}$: craving spikes — intense wanting with no expectation of pleasure. Because her recovery basin is still shallow relative to $W_{\text{sub}}$, the separatrix is close, and the cue supplies enough activation energy to push $\xi(t)$ across it: relapse = basin re-entry. Hysteresis means re-entry costs less than her first descent ($\Delta V^{\text{recovered}}_{\text{entry}} < \Delta V^{\text{naive}}_{\text{entry}}$), so within hours she is back at the addict's dose, not a beginner's — sensitization, not a fresh start. The framework's prescription is not condemnation but landscape surgery: medication-assisted treatment to raise the basin floor and shallow $W_{\text{sub}}$ (buprenorphine caps the supraphysiological excursion; naltrexone collapses the attractor by blocking its substrate), plus social reconnection and behavioral engagement to deepen $S^0_{\text{rec}}$ until its basin out-competes the wrong attractor.

Exercises

A patient washes for two hours after a contamination thought, feels relief, and reports the ritual "worked." Using the amplification equation $\hat{p}_{t+1} = \hat{p}_t + \alpha\,\hat{p}_t\,\mathbb{1}[\text{compulsion completed}]$, explain why the relief is simulated return, and predict what happens to $\theta_n$ and $C_n$ over the next month.
A clinician says "MAT just substitutes one addiction for another." Reframe the claim in landscape terms: what does buprenorphine do to $V(W_{\text{sub}})$ and to the inter-attractor gradient $\nabla\Phi$, and why does this differ from demanding willpower against an unmodified landscape?
(Open-ended) Both addiction and OCD are wrong attractors, yet ERP blocks simulated return to force genuine extinction of $\hat{p}$, while addiction recovery emphasizes constructing a new $S^0_{\text{rec}}$. Are these the same therapeutic move described twice, or two genuinely different operations on the landscape? Defend your answer using the framework's own primitives ($\Phi$, $\Delta V$, separatrix, DC5), and name one prediction your view makes that the rival view would not.

Sources

Addiction as Neurobiological Wrong Attractor: Allostatic Displacement, Craving, Relapse, and the Construction of a New Ground State (addiction_relapse.tex) — the false/new ground state thesis and relapse as wrong-attractor return.
Addiction as Wrong Attractor Deepening: Tolerance as Basin Steepening, Withdrawal as Failed Return, and Recovery as Landscape Reconstruction (addiction.tex, paper 193) — tolerance, hysteresis, MAT.
Obsessive-Compulsive Disorder as a Self-Amplifying Wrong Attractor Loop (ocd.tex) — simulated return, the amplification equation, threshold reduction and cost inflation.

All three belong to the Displacement Framework series and are archived live on Zenodo; the framework reference is DOI 10.5281/zenodo.20390438.