While I agree with the sentiment, the two questions aren't directly comparable, unfortunately.
There are all sorts of biases that this particular question brings up. I would guess that somewhere close to 100% of those questioned drove a car, while around 14% of the UK population smoke. So, naturally, we expect that people would object more strongly to something that they were on the receiving end of.

But, humans are extremely poor at judging risks, particularly when the likelihood is low, or results won't be seen for a long time period.

In another question, 61% of people agreed that risk was “a natural part of driving”, whereas just 31% agreed when “driving” was changed to “working”.

Again, not equivalent; accepting a risk that you have control over, is vastly different to having a risk imposed on you. Its why everyone accepts risks at home that they would never accept in the workplace.

Cognitive bias is an interesting topic. I see it on a daily basis working in health and safety. Perceived risk does indeed very much depend on a person's background, what they are used to, and the context in which they find themselves. It is a slow process to change peoples risk perceptions but by keeping on chipping away at it, and challenge the norms, we will hopefully improve health and safety outcomes long-term in all areas.

Another question this study poses is whether we would attempt to evaluate risk with such a limited amount of information on the hazard.

While the questions seem similar, the visualisation of the hazard will vary quite considerably.
Vehicles will almost always be outdoors while smokers may be indoors (proximity and ventilation).
The variable in whether a car stuck in traffic or is moving at speed is greater than that of a person sitting smoking or walking around. There's no mention of how many and how often.

A reminder that we need to get our walking shoes on and see the hazards for ourselves before attempting to define a level of risk.

The questions are loaded as Smoking is a more emotive subject and people have defined views whereas vehicle fumes are not front and center of the conversation.

I think the reality is that as a society we have been raised to expect, or accept a whole lot of driving risk especially that imposed on people outside of vehicles that would not be acceptable in most other situations. The expression for this is 'car brain' or motonormativity. The fact that we refuse to apply basic H&S principles to those who create the most risk, and tend to look to blame others, usually the most vulnerable. Cars are rewiring our brains to ignore all the bad stuff about driving

Agree - motonormativity leads us to ignore the harm caused by vehicles because we are all complicit as drivers (or even as passengers).
That's why a few years ago Safeguard began to include media reports of truck crashes resulting in death (usually to a car driver) in the work-related incidents we report in Safeguard Update newsletter.
Just to try to make the point - probably ineffectually - that trucks are driven by workers who are in control of potentially lethal mobile plant, and that "health & safety" should encompass all the issues faced by truck drivers rather than treating commercial driving as purely a road safety issue and therefore ignorable by H&S people.

A very interesting discussion topic. In my opinion “Risk” is poorly understood as a concept ( as I have iterated in past Safeguard articles) and misunderstanding continues to proliferate. There are some obscure (read - unhelpful) definitions, such as “the effect of uncertainty on objectives”. The simplest way to describe risk in our H&S environment is in the relationship between Hazard and Risk. [They are definitely not the same]. Hazard = cause or source of harm / damage. Risk = the probability of a harmful / damaging event / exposure taking place. Of the 3 core factors that comprise a risk assessment, Severity can be reasonably estimated, in most cases; Frequency of possible Exposure is more easily assessed; but, Likelihood or Probability of occurrence is often difficult to assess, particularly if there is no known previous occurrence. Data collected can indicate to us actual severity and exposures, and often, probability, e.g crash rates for trucks on the Kaikoura section of SH1. Because it is thought that people with good driving skills are operating these vehicles the problem is with the road design and “other” drivers. Motonormativity shifts our focus away from the driver because we are all of a similar mindset about driving. But technology such as in-cab video monitoring is changing perceptions about causation. It’s not just video of crazy driving by other road users captured by trucks but driver monitoring as well. Research into driver impairment conducted by Monash University/ Otago University several years ago found drivers suffered almost identical levels of impairment after 17 hrs of wakefulness as consumption of alcohol up to the legal driving limit. Now we’re focussing more on driver wellness / lifestyle / driver scheduling as driving hazards. Now controls are beginning to be focussed on reducing risk by reducing driving hours (exposure & probability), etc etc. Engineering controls (design) applied to the roading network help as well.