Psephologist · Professor, Strathclyde · President, British Polling Council · BBC elections analyst · Expert Watch · monthly scan
Why we watch him. Britain's most authoritative, most-quoted polling voice, now publicly staking out the sceptical-but-hybrid position on AI-simulated opinion versus real polling. His verdict is close to a market signal on whether "synthetic respondents" (Electric Twin, Aaru, Artificial Societies and similar) are treated as credible by the research establishment, which bears on the addressable market and defensibility of any AI-opinion-simulation startup.
Current headline view: polls remain useful but must be read as estimates with uncertainty, not precise forecasts; AI can be used in opinion research but "cannot dispense with all human testimony."
Latest 1-page summary (as at 9 Jul 2026)
Hybrid, not replacement, on AI opinion research. At the 27 Jan 2026 MRS debate he concluded AI is "to be used, but not that we can dispense with all human testimony."
Modelling is what pollsters already do. "Modelling the data is exactly what we do", rejecting the framing that synthetic modelling is new.
AI models drift from reality without fresh human data. Simulations decay unless continually refreshed against real survey input.
Individuals matter, not just aggregates. Research must still care what individuals actually think and do.
Polls are estimates, not predictions. Treat every poll with a margin of error; read the trend, not the headline number.
Transparency in the industry. As President of the British Polling Council he champions methodological disclosure and accountability.
Volatility caution. Repeatedly flags how soft and volatile current party support (e.g. Reform) is, cautioning against over-reading swings.
Credibility gap on AI polls. His scepticism aligns with findings that "silicon sampling" is generally not a reliable substitute for human respondents.
Notable reactions (operators & researchers)
High-signal reactions from operators or researchers (not commentators) to a specific argument of his. Challenges first. As at 9 Jul 2026 (baseline; entry below predates the standard monthly window and is included as the defining recent challenge).
Ben Warner (co-founder, Electric Twin; operator and ex-government data scientist) challenged Curtice's position at the MRS "Big Debate" on 27 Jan 2026, arguing traditional research is too slow and costly ("six weeks and £50,000"), that the say-do gap already undermines survey data, and that synthetic audiences give near-instant insight. Curtice's side prevailed on the vote, but Warner's challenge is the live counter-case to his hybrid stance. Source: Research Live, 27 Jan 2026.