Real world problems have too many pieces of evidence and perspectives that vary over time to easily calculate the impacts on probability with each revision of the evidence.   

The gambling industry has historically used odds because they are easier.   The bookmakers create an initial set of odds for each contest and then update the betting odds based on wagers.  The amateur gamblers have simple methods of estimating the odds.   The professional gamblers make their own set of odds and bet on the contests where the amateurs as group made the biggest mistake in estimating.   Today the gambling on sports, commodities, futures and stocks uses computers.  

The gambling that individuals do with their social media consumption does not yet have an app for updating the probabilities of truthfulness.   The dialog between an AI based app and the end users concerning the reasons for the assessment will be a lot less frustrating if both the AI and the users think about the likelihood of truthfulness in the same way. 

Best would be to use odds for likelihoods of 137 to 1 or better (~3 sigma) and the base 10 logarithm of odds for longer odds.  

A favorable real world using odds example from @3blue1brown on YouTube.  The medical test paradox, and redesigning Bayes' rule.

Disadvantages - 

• Grade school children might become more capable gamblers which will offend some people

• The change in odds concerning gambling addiction from this specific improved understanding is untested.

  • There's a risk which should be assessed and mitigated.

Load Weighted Odds Comparison

When there are options for which product to buy, action to take, strategy to use or conclusion to accept as true each product option can be broken into smaller criteria and evaluated individually.  Best practice is to begin by creating the list of criteria and then assign relative importance with a load factor.  When the product is a model of reality, a grand unified theory,  the load factors become odds of correctness. 

Generally, there are some criteria which are so important that the entire option is rejected based on a single criteria. If all possible options are rejected then one or more of the criteria need to be broken down into parts.