iiinova — where your digital self keeps evolving

James Winters

@replicatedtypo.bsky.social

1.3K followers 757 following 80 posts

Simulating sci-fi novels and calling it cultural evolution Centre for Culture and Evolution | Brunel University London | j-winters.github.io

Posts Replies Media

James Winters @replicatedtypo.bsky.social · 12/08/2026

My eclipse shades didn't arrive from Amazon, but they did supply the necessary equipment for my wife's bespoke Amazon pinhole camera

010

James Winters @replicatedtypo.bsky.social · 05/07/2026

So much for the wisdom of the crowd

110

James Winters @replicatedtypo.bsky.social · 21/05/2026

Back in the office after a fantastic CES-HBES conference. Big thanks to the organisers @matcharbonneau.bsky.social @anthrosarah.bsky.social @edseabright.bsky.social @zhgarfield.com & support team. Came away feeling motivated to get writing and researching! #CES2026 #HBES2026 @ces2026.bsky.social

0163

James Winters @replicatedtypo.bsky.social · 26/09/2025

And most of the time the rest of us wish you said nothing at all.

020

James Winters @replicatedtypo.bsky.social · 15/08/2025

Under the right conditions, this causes open-ended growth to accelerate. Why acceleration? As complexity grows → more resources → more iterations → higher chance of finding expansions → positive feedback. 🧵: 25/29

Selected simulation runs showing a range of dynamics. Coloured lines represent the complexity of technological systems (denoted with η) and search spaces (denoted with λ) for different simulation runs. Several simulation runs show accelerated growth in the complexity of technological systems. This includes: η=0.80, λ=0.80; η=0.95, λ=0.95; η=0.95, λ=0.99; η=0.99, λ=0.95. As one parameter or the other decreases, the dynamics slow down and show less potential for accelerated growth, e.g., η=0.99, λ=0.10; η=0.95, λ=0.05. And some simulation runs fail to exhibit any open-ended or accelerated growth, i.e., η=0.10, λ=0.99; η=0.99, λ=0.99.

110

James Winters @replicatedtypo.bsky.social · 15/08/2025

Open-ended evolution in bitw0r1d only emerges in a sweet spot where some stochasticity prevents the dynamics from settling into a stable equilibrium, while strong selection pressures helps maintain effectiveness. 🧵: 20/29

A heatmap showing the average complexity of technological systems for all simulation runs across η and λ at the end of the simulation. It shows the average complexity of is low for most regions of the parameter space, but it increases as η and λ approach 1.0. The regions with highest complexity are when η=0.95 and λ=0.80 and when η=0.90 and λ=0.95.

100

James Winters @replicatedtypo.bsky.social · 15/08/2025

Although absorbing barriers are widespread (~98% of sim runs), they are most prevalent when stochastic factors dominate (low η, low λ). Here, 𝕋 and 𝕊 drift independently, leading to low effectiveness and an inability to sustain long-term resource production. 🧵: 17/29

All simulation runs (coloured lines) showing the complexity for technological systems (y-axis) for η=0.01 and λ=0.01 over generations (x-axis). When η=λ=0.01, none of the runs manage to survive for the full 10,000 generations and are restricted to relatively simple technological systems.

100

James Winters @replicatedtypo.bsky.social · 15/08/2025

Missing from these models is any explicit representation of search spaces (𝕊). A key contribution of our model is to treat 𝕊 as culturally evolving entities alongside 𝕋. 🧵: 7/29

A diagram outlining the general dynamics of bitw0r1d. At each generation, a society needs to find effective mappings between 𝕋 and 𝕊. The effectiveness of this mapping as well as the complexity of 𝕋 and 𝕊 determine the amount of resources (R). These resources are then allocated to two processes: one which changes the technological systems of a society and another which changes the underlying search space. We assume both processes follow a two-step cultural evolutionary dynamic of generating and adopting changes. Our model varies the selection pressures on 𝕋 and 𝕊 by manipulating the extent to which the adoption of changes are stochastic or deterministic: η controls the selection pressures on 𝕋 and λ controls the selection pressures on 𝕊.

120

James Winters @replicatedtypo.bsky.social · 18/10/2024

New preprint with Niek Kerssies and Jose Segovia-Marting : "Connect-while-in-range: modelling the impact of spatial constraints on dynamic communication network structures" (arxiv.org/abs/2410.13527) #NetworkScience #SpatialModeling #NetworkTopology #AgentBasedModeling (thread: 1/9)

265