Interaction, not dissection: reframing AI interpretability
This paper argues that mechanistic interpretability should be understood as experimental interaction with models rather than static mapping of hidden mechanisms. It reframes what the field can legitimately claim to reveal.
From: European Journal for Philosophy of Science
The author offers a philosophical reinterpretation of mechanistic interpretability methods.
The paper suggests these methods can uncover behaviors and processes that emerge under intervention.
It does not supply empirical evidence about model experience, consciousness, or welfare.
Source: Springer ↗. This is a summary; the source is the authority.
This is a conceptual shift in how interpretability should be framed, not new evidence that AI models have experiences.
How we weigh evidence like this: the evidence levels.
- Mechanistic Interpretability Should Prioritize Feature Consistency in Sparse AutoencodersThis paper argues that sparse autoencoders should be judged not only by reconstruction quality and sparsity but also by whether the same fea…
- Mechanistic Interpretability for Neural NetworksThis is a broad survey of mechanistic interpretability, covering the goal of reverse-engineering internal algorithms in neural networks. It …
- Constructing Interpretable Features from Compositional Neuron Representations in Language ModelsThis paper proposes an alternative to standard sparse autoencoders by decomposing MLP activations with semi-nonnegative matrix factorization…
- Interpretability Transfer from Language to Vision via Sparse Autoencoder Transfer AlignmentThis work transfers labeled sparse-autoencoder features from language to vision inside a multimodal model. It uses those features to make vi…