This is a simple and straightforward guide to designing human-centered AI. The techniques mentioned herein were honed from our own experiences designing AI systems across industries and segments, for both consumers and enterprises. Our work in AI has taken us from finance to healthcare, from VR to photo-sharing, from asset management to predictive operations, and much more. Designing an AI to work within the messiness of the real world requires new frameworks, as novel challenges emerge within such dynamic and complex systems. While we cannot offer a step-by-step process that guarantees innovation (no innovation can come without experience and tinkering), this guide attempts to distill our own learnings into a reusable methodology.
We are not the first ones to utter the phrase ‘human-centered’, or even ‘human-centered AI’. Therefore, we do not take sole responsibility for the ideas presented throughout. However, we have worked to make them palatable within and beyond design circles, in the hope of inviting many different creative and ambitious people to contribute to the conversation.
A Useful Definition of AI
In essence, the easiest way to think of an Artificial Intelligence (AI) is as a computer program that takes a dataset as input (e.g. an Excel spreadsheet, a database, a file, etc.), and applies one of a wide variety of algorithms to compute correlations and relationships in that data. Therefore, at its core, AI is a set of techniques driven by statistics, sharing the same strengths and limitations.
AI decisions can range across types of information, contexts, and environments; some AIs are extremely narrow (e.g. deciding whether to recommend a certain movie to someone) and some can be mind-bendingly expansive (write a poem about anything). The less that a person has to encode hard ‘rules’ or ‘boundaries’ into the AI, the more general this AI is (and roughly speaking, the harder it becomes to build that AI).
Processing data to create an AI is called Machine Learning (ML). An algorithm processes massive quantities of data from which the AI is ‘trained’ to derive features and correlations by analyzing relationships within the data itself. For example, an algorithm that reads through home listings may create correlations between how many rooms a house has and the resulting price. As this field of AI has advanced, researchers have been able to create systems capable of interpreting much more complex correlations, including the correlations between pixels that make an image of a ‘cat’, or the correlations between words that make a semantically coherent story.