In reality the distinction between fixed and random effects often depends on the practical use and interpretation of parameter estimates. When the investigator is interested in comparing specific levels of a certain factors (let’s say amount of fertilizer for a plant or concentrate for a cow) then it is sensible to consider them as fixed effect. When a parameter is not of relevance for the analysis but rather a nuisance that we want to account for, more often than not we end up treating that effect as random.