Note that the actual volatility is directly affected by changes in knowledge. The implied volatility is directly controlled by investor expectations, and those are affected by expected events, so the implied volatility is indirectly affected by earnings reports. Actual volatility is a parameter, implied volatility is a measured statistic.
If either good news is certain, or bad news is certain, then volatility will not increase. The less certain the news is, the more volatility increases, regardless of whether the uncertainty is of the form "probably good news, but might be bad" or "probably bad news, but might be good".
However, "volatility" is a single parameter. It is used in models such as Black-Scholes that treat the spread as being characterized by a single parameter. Events like earnings reports tend to make Black-Scholes less valid, as Black-Scholes is justified by treating stock prices as a random walk that tends towards a normal distribution due to the Central Limit Theorem. When one event such as an earnings report is dominating a period, the Central Limit Theorem, and thus Black-Scholes, has reduced validity. While volatility just measures the variance of the spread, large events can make other parameters, such as skew and kurtosis, important considerations. In some cases, earning reports can create bimodal distributions of expected price.
As an example, suppose a stock has 90% probability of going up $1 and 10% probability of going down $9. Normally, a call and put options are roughly symmetrical on short time periods; a put option with a strike price $2 below the spot price costs about the same as a call option $2 above the spot price. But here, a put option $2 below the spot price is worth $0.70, while a call option $2 above the spot price is worthless. Just looking at the volatility, there would be no way to know this; while the relative value of puts and calls depends on whether it's "probably good news, but might be bad" or "probably bad news, but might be good", volatility does not.