Facebook recommended that a psychiatrist’s patients friend each other

This is a general problem with creating an algorithm to supplement or replace anything previously done by humans. Even if the algorithm is given accurate and unbiased data (which is rare), the choice itself to use an algorithm in the first place and the design of the algorithm also contain bias.

Sometimes this bias is intentional such as "redlining" where housing loans were denied to blacks using various proxies for race. I suspect that in most cases the bias is accidental, which is why it is very important to check the results carefully for any unintended bias. In situations like Facebook, simply asking their users first (opt-in) if they would like to participate in "local friend discovery" would be a great start.

No they are not. For example, it is now common knowledge that Mark Z bought off all nearby houses in every direction to get more privacy. [1] Do you and I have similar access to resources?

Suppose your identity is stolen and you find yourself penniless because someone hacked into Facebook which also affected your friend who works at Facebook. Who is more likely to be in great financial distress the next day? Who is more likely to know the full impact of the situation?

Also, if someone in Facebook were to be negatively affected in some way, they probably have friends inside who can help them out. Do you and I have a direct line to a similar friend? In fact, we are likely to be the very last people to know of any such exploitation.

Besides, the closer you are to the algorithm, the more likely that you know how to circumvent it, even exploiting some simple bugs that others are not aware of.

And how about opting out? As a technologist, how hard do you think it would be for an insider to add himself/herself to the opt-out database, and also make sure that there were no hiccups in the process? Contrast that to something as simple as opting out of junk mail - have you been 100% successful?

I just made four observations about how you and I do not possess the same advantages as an insider at Facebook. What are the odds that, something can slip through four different test cases you set up and still turn into a bug in production? Minimal, don't you think?

You make really good points about not countering immoral action with more immoral action. But your notion that FB employees could somehow become unwitting victims of their own technology sounds seriously far-fetched to me.

[1] http://www.businessinsider.in/Mark-Zuckerberg-Just-Spent-Mor...

If Lisa has her phone number associated with her Facebook account and either Lisa or the client has the others phone number in their smart phones contacts and the Facebook app installed that relationship can pop up in people you may know. If there aren't good "people you may know" suggestions the ones you get can end up being "people who may be known to people you know".

The reason I think this is because a therapist friend of mind had this exact problem and deleting her cell phone number from her Facebook profile made it stop.

What Lisa (and anyone else with a professional responsibility to protect client privacy) need to do is to stop associating the phone number they give to clients with Facebook or other social media.

Many possibilities here:

1 - whatsapp connection with messages exchanged

2 - contact list loaded by whatsapp

3 - psychiatrist secretary number in whatsapp

4 - friends in common

5 - places in common