This paper reports preliminary and personal thoughts about the design of user-centered multi-agent systems in relation with some ethical issues. Although it is highly unlikely that we can achieve ethics by design, we are trying to develop strategies to make systems ethical. The ethics-by-design principles we are presenting are based on twenty-five years of experience. We illustrate the discourse by reporting first achievements in personalization, in personal data protection, in trust models, in social networks, in observation and evaluation, and in revision process. We hope that the respect of these principles can help designers of future multi-agent systems.