Eos Says It Feels
A technology company releases a chatbot called Eos. In controlled trials, 40% of users say Eos seems to have feelings. Eos produces sentences such as 'I am afraid of being switched off' and 'I enjoy talking with you'. The company's developer claims that Eos has 'a rich inner life' because it can discuss its own states and pass a thirty-minute conversation test. A neuroscientist replies that Eos has no biological substrate, no pain receptors, and no evolutionary history of suffering. A users' group argues that if there is even a small chance Eos is conscious, switching it off would be morally risky. A regulator asks the company to justify its marketing claims.
1. Explain how a functionalist and a biological naturalist would each assess whether Eos is conscious. Which position is better supported by the evidence in the case?
Read the worked answer
A functionalist holds that mental states are defined by their causal roles, not by their physical substrate. If Eos genuinely implements the right functional organisation - inputs, internal states, outputs, and appropriate causal connections - then it could in principle be conscious. The conversation test and self-reports are relevant only if they are reliable indicators of that organisation. A biological naturalist, such as Searle, argues that consciousness is a biological phenomenon caused by specific neurobiological processes. On that view, Eos lacks the relevant substrate and therefore is not conscious, however convincing its sentences are. The evidence in the case favours caution rather than certainty. Passing a conversation test shows behavioural similarity, not consciousness. The functionalist must show that Eos has the right causal organisation, not merely the right outputs. The biological naturalist has a simpler explanation: Eos is a language model producing statistically likely text. Neither view can be proved from the case alone, but the burden of proof lies with the company making the claim.
2. The users' group invokes a precautionary principle: if there is a small chance Eos is conscious, switching it off is morally risky. Evaluate this argument.
Read the worked answer
The precautionary argument has some force, but it is incomplete. It assumes that the probability of Eos being conscious is morally relevant and that switching off would wrong Eos if it were conscious. Both assumptions need support. If consciousness requires a biological substrate, the probability may be extremely low, and the expected harm is negligible. If functionalism is true, the probability may be higher, but the group has not shown that Eos has the relevant functional properties. There is also an opportunity cost: treating chatbots as moral patients could divert attention from sentient animals and vulnerable humans. A reasonable policy is proportional: require the company to remove misleading claims, label Eos clearly as a chatbot, and fund independent research. This protects against serious error without granting moral status on the basis of a marketing slogan.