Analogue games run on the human wetware operating system (Mark Rosewater's Lesson #1). Akin to digital games, we must tradeoff 'rules fidelity' and 'runtime cost', and acknowledge different players have different performance characteristics and thresholds for rules complexity. Well known examples of 'fiddly' or nonintuitive rules in D&D 5E are bonus actions and casting levelled spells, subtle differences between ability (skill) checks and saving throws and attack rolls, and surprise and hiding and perception in combat.

If we want to try and simplify a task that we find is too complicated or unintuitive in play-testing, we could consider the seven principles from Don Norman's The Design of Everyday Things. He argues that objects afford (invite) certain interactions by their perceived and actual properties, and that the role of the designer is to adapt the design to try and meet user intuition, rather than providing additional explanations or training to overcome their inclination. Here are his principles (slightly adapted to reduce redundancy) and how we might apply them in imagination gaming.

Seven principles of simplifying tasks

1. Use knowledge in the world and in the head. There exist 3 conceptual models: the designer's, the user's, and the 'system image' (object as actually exists). The latter is the 'game text' in our context. An application of this include first-class design of character record and play sheets as they are the 'user interface,' and will strongly influence the player's model of how different attributes or procedures interact (are skills listed under ability scores or apart in a single list? are hit points listed as a small text field to erase-and-write or a larger field of tally or check boxes? are the constraints of your inventory/encumbrance system conveyed clearly like slots or does it need referencing of pounds and weight limits with rulebook in play?).
2. Simplify the structure of tasks. Conscious minds have limits on concentration, short and long term memory, arithmetic - and these vary significantly between individuals and occasions. Mental aids can help to overcome these limits: keeping arithmetic simple (comparison is easier than addition is easier than subtraction for most), quick reference tools on play sheets or inside cover of book for commonly-referenced details that are not necessarily worth player's rote learning (like Lamentations of the Flame Princess' equipment lists) sheets/inner cover, or automating the task (character or equipment packages or random tables rather than à la carte inviting analysis paralysis). Otherwise, we might change the nature of the task altogether, like supply rules or just-in-time equipment selection in Blades in the Dark.
3. Make things visible: bridge the gulfs of Execution and Evaluation. Transparency of state before (Execution) and after (Evaluation) each interaction helps develop an effective user model. This relates more to play culture and effective table talk, and might include closed-loop communication around dice rolls to ensure we all have the same conceptual model of what's at stake, and how the situation is subsequently changed.
4. Get the mappings right. How do intentions map to actions, actions to effects, and is the system state well conveyed by the senses. On a play culture level this is encapsulated by the ICI Doctrine, but it's worth additional emphasis that our channel for communication (receptive and expression) in play has far narrower bandwidth than our full set of senses in meatspace. Therefore clear and explicit communication of intents, actions, risks, and results is paramount and may need to be somewhat 'extrasensory.'
5. Exploit the power of constraints, both natural and artificial. Our aim is to afford the user the single right way to interact, and constrain alternatives that cannot be accomodate. I see relevance to randomised character generation procedures (ensuring the full gamut of outcomes is playable without recourse to arbitrary re-rolling of 'unplayable' results) as well as class and level games nullifying one of their prime advantages by providing multi-classing that tends to introduce many character builds with wildly varying complexity and potency.
6. Design for error. All systems require fault-tolerance, and to err is human. Expect that players will forget, misremember, misapply, or with foolish intention change the rules - so normalise reflective play (such as Stars & Wishes and debrief of rulings to firm them up for future sessions). Avoid 'gotcha' moments by ensuring a shared understanding before committing to an action, and I personally think all games benefit from a form of Script Change. All games are conversations (not just the 'story games'), and we have available all our conversational and interpersonal techniques to have hot, sweaty, but safe discourse and roleplaying.
7. When all else fails, standardize. In our case, if you must use a mechanism that is counterintuitive despite attempting the above, aim to replicate something common and likely already known to the user. Although I have found upon introducing new players to D&D 5E that the division of ability score and modifier is not intuitive, given it is the market leader by far, other games can get away with keeping this if desired. The corollary is that rules novelty for the sake of it poses a barrier to entry, and will likely diminish the audience willing to persist all the way through to actually getting it to the table.

Tackling those bugbears

Back to the initial examples from D&D 5E - I'll propose one solution for each based on the above, though an infinite solution plane for each surely exists.

• Bonus actions and casting levelled spells. Simplify the structure - remove the restriction and permit both standard and bonus action levelled spell casting, and adjust specific spell descriptions to resolve any breaking interactions (there are surprisingly few to my understanding).
• Subtle differences between ability (skill) checks and saving throws and attack rolls. This is handled well again by simplifying the structure in Black Hack/Oddlikes by centralising around a combined ability check/saving throw mechanism. Yet an alternative is exploiting the power of constraints and influencing the conceptual models: making each check more different helps reduce confusion. Games like Beyond the Wall and Kevin Crawford's Without Number series exemplify this.
• Surprise and hiding and perception in combat. This is the hard one, and I'll make a somewhat contentious proposal: remove most of the rules text on these and leave it up to GM fiat (at a stretch, this is get the mappings right). During play outside of combat the game provides considerable latitude to the GM, as do most games, as it speaks to the fundamental challenge of conveying a richly imagined shared world of sensory details. All the oddities about two people in darkness or blinded attacking each other normally, or if surprise requires all or just any threat to not be perceived, would be easily ruled on a per case basis and the elision (abstraction) these rules provide is ultimately unhelpful.

Sticky wicket

Of course as games are in part defined by challenge, we may sometimes deliberately avoid simplification, or even deliberately complicating the matter by applying the inverse of the above techniques. This was famously demonstrated in the design of D&D 3E being influenced by Magic: The Gathering by what was coined Ivory Tower Game Design. In summary, that the designer lays out the mechanics but holds back from explaining the whys and hows, and this was manifest in deliberately providing better or worse player options to permit discovery and system mastery. If your target audience are inveterate grognards who have been around the traps, then these inverted techniques may provide pleasing results!

What are other examples of simplifcation? How might you apply these principles in designing or preparing for your own games? What is your tolerance for complexity, and do you sometimes partake of 'Ivory Tower Game Design'?