Role:
Senior Product Designer
Company:
DeepL
Key responsibilities:
AI product design
Design strategy
Synopsis:
AI-generated translations may lack contextual accuracy. As the designer driving Clarify, I designed the feature to bridge this gap by allowing users and the model to collaborate to achieve greater precision, clarity, and contextual accuracy.
Clarify: Designing User<>AI Collaboration at DeepL
Product overview
DeepL is the world's most accurate translator. With over 10 million daily users, DeepL Translator is known for its exceptional accuracy and natural fluency. Fast, reliable, and intuitive, DeepL is a go-to solution for high-quality, context-aware translations.
Problem space
Every translation makes assumptions and traditional machine translation lacks necessary user input required for greater precision, adaptability, and contextual understanding. The ambiguities such as those in gender, idioms, or specialised terms can result in confusing or misleading translations, especially for non-expert users.
Opportunities
Users need full proficiency in target language to refine the translation
Without expert knowledge in the target language, users struggle to grasp the quality of the translation, resulting in the lack of confidence in utilising the translation in their multilingual communication
AI translation makes assumptions without asking users for clarification
Users rely on machine translation but often find that ambiguous phrases go unnoticed, leading to misinterpretations
Editing translation can be costly and cumbersome
Manually adjusting translations or rewriting phrases to avoid ambiguity adds extra steps, making the experience frustrating and inefficient
Users in specialised fields lack terminology support
Translating industry-specific terms such as those in the legal, medical, or technical industries require professionals to cross-check and correct translations, increasing the cost and time for business to achieve their goals
Impact
Signal scanning
Through scanning the pool of customer support tickets and the interview notes, we identified that no matter how accurate the machine produces, it will always lack contextual that only human users can provide. These insights helped establish the need for a system that could involve users in clarifying intent, rather than relying solely on AI guesses.
Conceptualise
While initial work on training the model began, I focused on analysing patterns in translation errors.
Some of the key categories are:
Gender
Idioms
Formatting
Culturally specific terms
Mapping these categories allowed us to define the types of clarifications that would provide the greatest value to users.
Prototyping & user testing
I developed early prototypes that prompted users for clarifying questions when ambiguities were detected. We conducted customer interviews to collect early feedback and refined the experience based on their feedback.
Internal release
Before a public release, we conducted an internal launch to gather insights on:
Usability: Was the feature intuitive and non-disruptive?
Value: Did Clarify improve translation accuracy and was the effort required justifiable?
Scalability: Could the AI model efficiently handle a range of clarifications without overloading users?
Multiple iterations were tested internally and with key stakeholders, refining the experience based on feedback.
5. Experiment design & release
We identified key success metrics and create an experimentation plan to measure the impact for the launch.
The experiment was designed to track both quantitative metrics, such as engagement rates, and qualitative feedback, assessing how intuitive and helpful users found the feature through the survey.
First release
Next steps
Monitor post-launch metrics & user feedback – Continuously track engagement, error rates, and qualitative user feedback to identify areas for improvement
Collaborate with ML scientists to optimise AI models – Refine the AI’s ability to detect ambiguity and improve contextual recommendations
Iterate on UI/UX for a more effective and time-saving userflow – Address frictions to make the interactions more intuitive and integrated into workflows.
Scale to a broader user base – Gradually expand availability to more users and more languages
I am more than a sum of pixels.
I love facilitating collaboration, understanding different points of view and taming chaos.
When I’m not designing, you’ll find me surrounded by plants, expanding my reading list, or experimenting with new recipes in the kitchen.
Recently, I've been exploring alternative design approaches. Need book recommendations or just fancy a chat? Hit me up anytime!
See more: