Lingo
← Back to Neuroscience

Attentional Drift Diffusion Model (aDDM)

Overview

The Attentional Drift Diffusion Model (aDDM), introduced by Krajbich et al. (2010), is an extension of the standard Drift Diffusion Model that explicitly incorporates visual attention (eye movements or fixations) into the evidence accumulation process.

In traditional models, the drift rate is constant. In the aDDM, the drift rate changes dynamically within a single trial depending on where the subject is currently looking. Specifically, the model assumes that attention biases the accumulation process in favor of the attended item.

The Discount Parameter (θ)

The core innovation of the aDDM is the attentional discount factor, θ (theta), which ranges from 0 to 1.

  • When looking at the Left item, the drift rate is proportional to Value_Left - (θ * Value_Right).
  • When looking at the Right item, the drift rate is proportional to (θ * Value_Left) - Value_Right.

If θ = 1, attention has no effect (the model reduces to a standard DDM). If θ < 1, the value of the unattended item is discounted, creating a systematic bias toward choosing the item that is looked at longer.

Empirical Success

The aDDM successfully explains several widespread empirical phenomena in consumer choice and neuroeconomics:

  • Gaze Bias: People are more likely to choose an item they spend more time looking at, even if it has a slightly lower objective value.
  • Last Fixation Bias: The chosen item is highly likely to be the item that was fixated immediately prior to the decision.

References

  1. Krajbich, Armel, & Rangel (2010) — Nature Neuroscience— Visual fixations and the computation and comparison of value in simple choice.

Examples