According to the latest ABC News/Washington Post poll, a majority of Americans disapprove of the FBI’s recommendation not to charge Hillary Clinton with a crime over her handling of emails while she was secretary of state.

But the people who designed the poll chose not to ask about the issue in an objective way. Instead, they designed a tendentious question that leads a respondent to be more critical than supportive on Clinton’s emails.

Here is one possible way to ask about the issue in an objective way, to avoid leading the respondents in one direction or another:

Q1: How much have you heard about the FBI’s recent decision about Hillary Clinton’s use of emails while secretary of state – a great deal, a moderate amount, not much, or nothing at all?

2: The FBI Director announced that he recommended against prosecuting Clinton for her use of emails while secretary of state. Do you approve or disapprove of that decision?

Instead, this is the question that the ABC/WP included:

“As you may have heard, FBI Director James Comey has recommended NOT charging Hillary Clinton for her use of personal email while secretary of state, saying she did not have any criminal intent. He also said Clinton was ‘extremely careless’ in her handling of classified information in her personal email. Do you approve or disapprove of Comey’s recommendation that Clinton should NOT be charged with a crime?”

Note that the poll makes no attempt to find out how many people are even aware of the issue before feeding respondents information about it. Yet, discovering how many people are engaged enough to have an opinion and how many are simply not paying attention is a critical part of any poll that purports to measure what the public is thinking.

Instead, the poll feeds its respondents information, and then asks a forced-choice question (which provides no explicit “don’t know” option), and thus gives the impression that virtually all Americans have an opinion about the matter. After being given the information, just 9% of the sample volunteered no opinion.

And the information the poll provided shades the issue in a way that is negative to Clinton. The question does note that Comey said Clinton “did not have any criminal intent.” But then the question stresses that Clinton “was extremely careless” in handling the email. What the poll did not tell the respondents is that Comey also said “no reasonable prosecutor” would bring charges against anyone based on the evidence he had, and that Clinton was not being given special consideration.

The problem with a pollster giving any information to respondents is that the information is necessarily limited. And in deciding which information to give and which to withhold, the pollster is almost certainly going to provide a biased picture.

The only objective solution is to provide no information, as in the example provided above.

After giving the respondents limited information about Comey’s decision, stressing the “extremely careless” comment, ABC/WP then asks two follow-up questions – whether the outcome of the issue has made the respondents more or less likely to support Clinton, and does the issue make respondents worry how Clinton might handle her responsibilities as president if elected.

Not a surprise – after being primed with negatively biased information, respondents are more likely to give negative than positive answers to each of the questions.

It’s quite possible that even an objective poll would have shown more people with negative than positive views of the FBI’s decision. We don’t know. But the ABC/WP priming of respondents in a negative direction is an important lesson in how not to ask a question – unless the goal is to manufacture the semblance of a “public opinion” in a given direction.