You're working with an AI development team. How can you give feedback that they'll actually listen to?

??Grasping project goals is crucial before offering feedback. It ensures relevance and constructiveness. ?? Project Vision Understanding the AI's purpose guides relevant feedback, aligning with project goals. ?? Metrics Insight Knowledge of performance metrics is vital for targeted, useful feedback. ?? Constraints Awareness Awareness of limitations helps avoid unrealistic suggestions, fostering feasible improvements. ?? Your feedback becomes a catalyst for progress when grounded in the project's foundational goals and realities.

Before offering feedback to an AI development team, ensure you fully understand the project's goals, such as enhancing an AI system's ability to predict stock market trends. Recognize the constraints, like data availability or regulatory compliance, and the technologies employed, whether deep learning or statistical models. This comprehension not only demonstrates your respect for the team's efforts but also makes your feedback pertinent and actionable. For instance, recommending improvements in the model's data preprocessing steps is more valuable when it aligns with the overarching goal of increasing prediction accuracy and acknowledges existing limitations.

Incorporate practices from Amazon, TikTok, and others such as memo/document reading during meetings. Engage in a Two-Way Conversation where feedback is a dialogue, not a monologue. Encourage the team to share their thoughts, and reasons behind decisions, or actions. Try and offer suggestions for improvement or solutions to specific issues.

Understanding the AI project's goals is crucial for providing relevant feedback. It ensures that suggestions are not only technically feasible but also strategically aligned with the project's objectives. Metrics are essential for gauging success and should be well-understood to offer insights that could improve performance. Recognizing data and algorithmic constraints helps in offering realistic and implementable solutions, while understanding the environment can lead to feedback that enhances the AI system's robustness and adaptability.

Give specific examples (like princess stories for girls). Show how it hurts customers (limits options) and damages the company (bad reviews, lost sales). Mention solutions (broader story themes) and the risk of fast backlash. They'll listen to protect their product and reputation.

In AI development, specificity in feedback is crucial for actionable improvements. For instance, stating that "the AI's image recognition feature misidentifies cats as dogs 15% of the time" is more helpful than saying "the AI is inaccurate." This level of detail enables developers to pinpoint issues and assess the AI's performance against its intended goals. Additionally, providing concrete examples or data helps in creating a shared understanding of the problem and facilitates a more focused approach to troubleshooting and optimization.

This is an area where the 7-38-55 rule shines. In his highly-cited study Professor Albert Mehrabian demonstrated that 7% of a message is based on words, 38% from tone and 55% comes from body language. This is why feedback is better when given in person or at least over a video call. The goal of giving feedback is to get the message across without forcing the receiver into a defensive mindset. After all, feedback is given for the benefit of the receiver. The best feedback is worthless if the receiver doesn't take it on board. So, framing how you say it is particularly important. Asking open-ended questions that make the receiver rethink their position is the best way to achieve this. It is important to close out with a commitment to action.

Provide specific examples of the issue or behavior you're addressing. If it's about the AI's performance, point out exact instances where it failed or succeeded. Use data, logs, or user feedback to support your points. This makes your feedback less about personal opinion and more about tangible outcomes.

First and foremost, be a good listener: This is essential. Effective feedback starts with understanding the context and perspectives of those to whom you're providing feedback. Prefer quantitative details over qualitative: Providing specific examples and metrics can make feedback more actionable and clear, avoiding ambiguity or confusion. Cross-examine the team's philosophy: This encourages critical thinking and ensures that feedback is aligned with the team's goals and approach. Analogies and examples: Using everyday examples can make complex concepts more accessible and easier to grasp for the team. Moreover, ensure feedback is timely and a dialogue, not a one-way critique. Keep the tone constructive.

Ditch vague comments. Instead of saying "The AI is not user-friendly," offer specifics: "The initial setup process lacks clear instructions. Providing a step-by-step guide with screenshots would improve usability."