Language and cognition, MIT's Olivia Honeycutt study on framing

MIT senior Olivia Honeycutt presented research on May 1, 2026 showing that small differences in wording and question framing systematically shift people's beliefs and policy preferences. Her senior thesis uses a series of controlled online experiments that vary word choice, numeric formats and metaphorical frames across multiple topical vignettes to measure how communication shapes judgment.

Method and core findings

Honeycutt designed paired survey vignettes that differed only in one dimension of language. Across those pairs she varied items such as verb choice, label selection, the presence or absence of a metaphor, and how numbers were expressed. Respondents completed randomized assignments to versions of the same prompt so the study isolates the effect of linguistic choices from underlying respondent characteristics.

The experiments span several issue areas, with each experiment treated as a replication of the same basic manipulation strategy. Honeycutt reports consistent directional effects: alternative wording produced measurable shifts in self-reported beliefs and in expressed support for policy options. Effects size and direction varied by topic and by the type of linguistic change. For example, numeric framing and whether a situation was described using an agentive verb versus a passive construction produced different patterns of change than simple label swaps.

The study includes attention checks and balance tests to ensure randomization worked. It also reports subgroup comparisons that show framing effects are not uniform. Some demographic groups and attitudinal clusters proved more susceptible to particular linguistic manipulations, while others showed relative stability across frames.

Honeycutt situates the experiments within a broader literature on framing effects, question wording and survey methodology. The thesis draws on prior work in political science, cognitive psychology and communication studies, and it emphasizes both replication across contexts and the boundary conditions that limit generalization.

Context and implications

The findings reinforce a persistent lesson for researchers, pollsters and communicators: wording matters. Honeycutt moves beyond demonstration to map which kinds of linguistic shifts tend to produce the largest movement in responses and which produce little or inconsistent change. That mapping matters for anyone who designs public opinion instruments, crafts policy messaging or interprets survey-based evidence.

Honeycutt also flags practical considerations. Simple edits to a question can change measured public support for a policy enough to affect headline results, yet the same edits may have little effect in deliberative settings where respondents receive more context. The thesis calls for clearer reporting standards about question wording and recommends routine robustness checks when survey results are used to make policy claims.

Why it matters

Honeycutt's work underscores that communication choices can shape both measurement and the underlying attitudes they aim to capture, shifting how researchers and officials should read survey numbers. For poll designers, campaign strategists and policy analysts, the study demonstrates the need to treat question wording as an active factor rather than a neutral container. The mapping of which linguistic moves matter most also offers a practical toolkit for improving the reliability of survey-based evidence and for understanding when public opinion is deeply held versus when it is malleable.