What are the best practices for designing algorithms with end-user feedback?

The most important thing in my opinion is the degree of freedom you want to give to your end users. Adapting to your end user when designing an algorithm is as important as adapting to your audience when public speaking. You clearly need to understand their problem, KPIs and result they want before even starting coding a single line. Often they will think the tool you’ll give them has not limitations and will not understand if they discover it by themselves. Limitations need to be clearly defined !

I have to disagree on this one. Before even starting modeling, you often have to make sure you’re not reinventing the wheel, most of the time you’ll find some really good solutions already existing that can represent a super quick win for the final stakeholders or that can help validate you are going the right direction and solving the right problem. Exchanging with the end users at high frequency is the key. Diving deep directly into modeling can make you lose a lot of time, be careful !

When the likes of NP-Hard problems are solved, out-of-the-box optimization solvers fail to provide a good solution, let alone an optimal one. In such cases, incorporating end-user feedback into algorithm design might be very useful. Few pointers to consider include: - preference for a subset of decision variables - restrictions on a subset of decision variables - limits on variables and certain combination of variables Moreover, user feedback could be utilized to decompose the problem into meaningful smaller chunks, which in turn could be solved sequentially.

1. Engage decision-makers early to align with their business metrics and needs. 2. Use simplicity and real-world relatability when explaining algorithms. Welcome user feedback to improve. 3. Post-deployment, leverage surveys and metrics to improve algorithms via feedback loops. 4. Validate through A/B testing, quantifying performance gains versus targets. 5. Visualize flows for transparency, empowering stakeholders via understanding of optimization trade-offs. 6. Streamline enhancement requests through a quantified, criteria-based process. 7. Compare new versions to baselines, ensuring consistent metric improvements. Key is an iterative, collaborative methodology centered on user needs best positions OR algorithms for impact.

Algorithm design embraces user feedback. Actively engaging users via surveys and interviews, recent project feedback reshaped our algorithm, morphing it from a solution to a user-centric experience. A responsive design process acknowledges evolving user needs.Clearly communicating how user feedback influences decisions builds rapport. Sharing success stories, like when user input streamlined a recommendation algorithm, reinforces the value of their contributions. In the dynamic realm of algorithm design, user feedback isn't merely a checkpoint; it's the compass guiding algorithms toward meaningful, user-friendly outcomes.