This article was originally posted on Campaigns & Elections here.

It’s well-known to the scientific community, if not your parents, that cognitive biases impact the decisions we think we’re making based on rational choice.

That investment you didn’t make because you didn’t want to risk losing your money? That’s called loss aversion. Do you remember your grades in college as being better than they actually were? That’s an example of choice-supportive bias. And when you skipped school because everyone else did that one day? That was the bandwagon effect, which means that yes if everyone jumped off a bridge you might jump as well. Sorry, mom.

The argument dilution effect is another cognitive bias that guides our decision making. This effect explains that when we’re making judgments based on a lot of information — choosing medicine, making a large purchase, or voting — the details of all that information starts to average out. Whether relevant or irrelevant, considering an onslaught of less-relevant details right alongside the most important information turns it all into white noise, if you will. We’d like to believe we’re making deliberate, informed decisions that weigh information soundly, but we’re really buying a car as much because of its color as its resale value.

A good illustration of argument dilution came in a study in 2017 in Nature Human Behaviour that examined direct-to-consumer advertising of pharmaceutical drugs, specifically the comically long list of side effects that’s in every ad. The study showed that seeing more medication side effects makes people evaluate drugs as less harmful because they dilute the severity of each. In this study, patients were more likely to favor a drug when its minor side effects were mentioned alongside more major side effects, compared to just mentioning the major side effects. Turns out, this well-intentioned effort to inform patients ends up having its own side effect.

Other studies have shown a similar effect in assessing the guilt of suspects on trial or how consumers make judgments about brands, but no one’s looked into a similar effect in decisions about candidates for elected office.

Voters are bombarded with a lot of information about candidates, much of it arguably irrelevant to their suitability for office. There have been studies about how political messages via different media—direct mail versus TV ads versus volunteers knocking on doors, for example—affects voter behavior.

Others have measured how the tone, accuracy, and relevance of those messages impact the choices voters. But how practitioners focus their political messages is largely guided not by science but by the assumption that a focused, repeated, and simple message is best.

In September, we conducted an online experiment with more than 1,200 US voters showing two different sets of voters different tweets by the same politician.

Each tweet began with a sentence you might see in a typical politician’s tweet, such as “If re-elected as your Governor, I will bring high-quality jobs to our state and ensure that we have the best schools and colleges.” To the tweet one of the groups read, we appended that standard message with a largely irrelevant sentence “I will never be even a minute later to work. I will take turns getting coffee and breakfast for my assistant.”

When it came to qualification, commitment to winning, and likelihood of representing “people like you” each group rated the candidate more or less equally. But the diluted tweet with irrelevant statements about coffee and breakfast made respondents rate the candidate as about 20 percent less likely to win, and they were 10 percent less likely to watch or read about the candidate.

This suggests that voters may be more convinced by shorter messages that stay on message and that the perception of electability could be a factor in whether people learn more about the candidate—all of which conforms to conventional intuition.

Remarkably, the effects were reversed when we experimented with similarly appended statements geared at attacking their opponent. One group saw the tweet: “My opponent in the upcoming election has no experience in public office and has never served in our Armed Forces,” and the second group saw that with the following added to the end: “He admits to rarely if ever recycling his newspapers and used plastic bottles. He even failed his history exams—twice!”

You’d be excused for expecting the longer diluted attack including struggles with exams to strike the voter as irrelevant and, consequently, off-putting, and the simpler attack limited to the lack of military service as more persuasive, but no. The “kitchen sink” tweet (military service plus recycling and history exams) made people think the opponent was about 15 percent less likely to win and an equal percentage less likely to represent “people like you.”

These findings have implications for widely held assumptions about the 2016 presidential campaign.

We analyzed the last 100 tweets sent by Donald Trump and Hillary Clinton leading up to the last presidential election. Perhaps contrary to conventional wisdom, Clinton took the kitchen sink approach, whereas Trump was much more on message, repeatedly calling her corrupt.

Additionally, out of Clinton’s 100 tweets posted leading up to the election, 78 were messages attacking Trump, while he only tweeted about her 22 times. The conventional wisdom is that his strategy worked and hers backfired. These assumptions, which are driving much of the political theorizing ahead of the 2020 campaign, aren’t supported by our study or by her actual success with the popular vote, which suggests she may have made the more effective attacks, but failed in her demographic strategy at the local level.

Of course, more study is needed to confirm and expand these findings, but our new experiment begs campaign strategists a rethink — not a complete trashing — of the political playbook. Long story short, when you’re selling yourself, keep it simple and direct or they won’t think you can win and may not be open to learning more. But when you’re attacking your opponent, a more diluted argument, or “kitchen sink” approach, seems to do more damage.

Lilly Kofler is the U.S. Director of Behavioral Science at Hill+Knowlton Strategies, an international public relations and public affairs consultancy. Kofler has both studied and practiced behavioral science in both academia and communications for nearly a decade and holds a Master’s degree from the University of Chicago.