Confirmation bias, scaled up
An AI model trained on a narrow run of market history can hand an investor back their own prior in polished prose. We check whether the tool lets you see the counter-evidence or just flatters the view you arrived with.
AI in investment · behavioural biases & decision support
Kinetic Console Journal
A free editorial journal that checks how artificial intelligence surfaces behavioural biases in investment decisions — and where it quietly introduces new ones. Read the articles, then send an inquiry if you want to discuss what the evidence does and does not support.
What this journal checks
Three biases worth checking before you trust an AI-assisted trade
Recency and availability drift
Recent volatility weighs heavily on both human and model. Our articles trace how a model's window length changes the recommendation, and why a confident answer this quarter can be a fragile one next quarter.
Automation bias
The cleanest interface can read as certainty. We separate the parts of a decision the model is actually good at — pattern sifting across thousands of filings — from the parts it merely states with confidence.
How an article is built
The checklist every piece runs through before it is published
- 1
State the bias in plain words
Each article opens by naming the behavioural bias under review — loss aversion, anchoring, herding — and the specific investment decision it distorts, so the claim is legible before any model output appears.
- 2
Show what the model was given
We publish the prompt shape and the data window, not just the answer. A recommendation pulled from five years of a single sector is not the same animal as one drawn across three cycles.
- 3
Mark the confidence, then mark it again
Where a model returns a probability, we report the calibration question, not just the number. A 70% that behaves like a 50% is the whole story.
- 4
Separate decision support from decision making
The closing section states plainly what the article will and will not help you do. It supports thinking; it does not execute trades, size positions, or give personal advice.
Who this is for — and who it is not for
Read it if you weigh AI input in your own investment thinking
This journal is written for individual investors and analysts who already use, or are considering, AI tools as one input into their own decisions — and who want a calmer account of where those tools help and where they inherit human biases at scale.
It is not for anyone seeking a managed portfolio, a trade signal, or personalised investment advice. We sell nothing here: no subscriptions, no paid reports, no execution, no deposits. The only action you can take on this site is to read, and to send an inquiry if you want to discuss an article.
