Lame - you presented the poll, and you are now defending it. So, take responsibility for it.
This poll question shows a complete lack of understanding of polling methodology - or perhaps a complete lack of interest in achieving un-influenced, theoretically "fair" results. Great effort is normally expended by serious pollsters to conceal their own preferences or expectations from respondents. In this instance, the poll question embodies an obvious bias on the part of the questioner(s), and would suffer from high susceptibility to skewed results.
The question presumes a) that US actions were "unilateral," and b) that they were "against Iraq." A less biased, more accurate formulation might simply have dropped the word "unilateral" and have substituted the preposition "in" for "against." A reasonable respondent might, for instance, object to those aspects of the action that someone (such as the questioner) might consider "unilateral," but feel positive about whatever else about the operation had a multilateral character. Considering that the question was apparently being asked of Australians - citizens of country allied with the US - the reference to unilateralism is extremely suspect: It implies to the minimally well-informed respondent that his or her country has somehow been tricked or forced into participating in a US adventure. Another alternative might have referred to actions "against Saddam Hussein" or "against the regime of Saddam Hussein" - and would very likely have results more favorable to the American side.
In short, the question already makes the pollsters' desired and expected responses clear. The choices for respondents confirm and extend these biases:
Respondents were not given the option to respond that the action had increased their "sympathy." Even where such questions are not considered likely to be chosen, the presentation of the full range of logical alternatives is thought by pollsters and market researchers to help draw a more honest response from any significant opinion sampling. Excluding this choice further underlines the pollster's obvious belief that the answer to which they gave first position is the "correct" one.
Someone interested in receiving a fair, objective opinion sampling would also strive to make the choices as similar to each other as possible in phraseology. In this poll, the "correct" answer is also the only choice that offers an affirmative, yet non-absolute statement. Both "2" and "3" require the respondent to adopt absolute and negative positions - of having "views" that have not changed at all, or of "never" having had any sympathy. Furthermore, while "1" and "3" refer specifically to the issue of "sympathy for the Americans," "2" stands apart, in the passive rather than active voice, and makes no reference to the Americans, and substituting the notion of "changed.. views."
It's worth noting that the use of the word "sympathy" is also tendentious, though in a way that bears more on the potential usefulness of the poll rather than on its construction as above: Somebody who fully supported American actions might feel less "sympathy," in the sense of feeling less pity for Americans. It could be argued that the poll merely inquires about a relatively narrow issue - the effect of the Iraqi operations on 9/11-related "sympathy." It should be obvious that, as a defense of the poll's methodology, this argument would also militate against making any broader conclusions about the feelings of respondents about the war in Iraq per se: It's quite possible for someone to feel uncertain or even very favorable about the war, yet also note a reduction in "sympathy" for the war's victor. It would be natural even and especially for American war-supporters to feel less sorry for themselves in the aftermath of self-assertion and victory on the world stage.
The poll does a much better job of exposing the questioners' biases than it does of drawing meaningful results.