Observation research.

Direct (Reactive) Observation

In direct observations, people know that you are watching them. The only danger is that they are reacting to you. As stated earlier, there is a concern that individuals will change their actions rather than showing you what they're really like. This is not necessarily bad, however. For example, the contrived behavior may reveal aspects of social desirability, how they feel about sharing their feelings in front of others, or privacy in a relationship. Even the most contrived behavior is difficult to maintain over time. A long term observational study will often catch a glimpse of the natural behavior. Other problems concern the generalizability of findings. The sample of individuals may not be representative of the population or the behaviors observed are not representative of the individual (you caught the person on a bad day). Again, long-term observational studies will often overcome the problem of external validity.

What about ethical problems?
Ethically, people see you, they know you are watching them (sounds spooky, I know) and they can ask you to stop.

Unobtrusive Observation.

Unobtrusive measures involves any method for studying behavior where individuals do NOT know they are being observed. Here, there is not the concern that the observer may change the subject's behavior. When conducting unobtrusive observations, issues of validity need to be considered. Numerous observations of a representative sample need to take place in order to generalize the findings. This is especially difficult when looking at a particular group. Many groups posses unique characteristics which make them interesting studies. Hence, often such findings are not strong in external validity. Also, replication is difficult when using non-conventional measures (non-conventional meaning unobtrusive observation). Observations of a very specific behaviors are difficult to replicate in studies especially if the researcher is a group participant. The main problem with unobtrusive measures, however, is ethical. Issues involving informed consent and invasion of privacy are paramount here. An institutional review board may frown upon your study if it is not really necessary for you not to inform your subjects.