SoylentNews is people
SoylentNews
Algorithms tell you how to vote. Algorithms can revoke your driver’s license and terminate your disability benefits. Algorithms predict crimes. Algorithms ensured you didn’t hear about #FreddieGray on Twitter. Algorithms are everywhere, and, to hear critics, they are trouble. What’s the problem? Critics allege that algorithms are opaque, automatic, emotionless, and impersonal, and that they separate decision-makers from the consequences of their actions. Algorithms cannot appreciate the context of structural discrimination, are trained on flawed datasets, and are ruining lives everywhere. There needs to be algorithmic accountability. Otherwise, who is to blame when a computational process suddenly deprives someone of his or her rights and livelihood?
But at heart, criticism of algorithmic decision-making makes an age-old argument about impersonal, automatic corporate and government bureaucracy. The machine like bureaucracy has simply become the machine. Instead of a quest for accountability, much of the rhetoric and discourse about algorithms amounts to a surrender—an unwillingness to fight the ideas and bureaucratic logic driving the algorithms that critics find so creepy and problematic. Algorithmic transparency and accountability can only be achieved if critics understand that transparency (no modifier is needed) is the issue. If the problem is that a bureaucratic system is impersonal, unaccountable, creepy, and has a flawed or biased decision criteria, then why fetishize and render mysterious the mere mechanical instrument of the system’s will ?
(Score: 2) by kurenai.tsubasa on Friday May 22 2015, @04:08PM
Well, this is the larger problem of bureaucracy, but I think your situation is more hopeful than what TFS mentioned.
The clerk has no more power than the form on the screen. In fact, I wouldn't be surprised if the clerk was using that same form. The question I would have is why didn't she escalate the issue to her superiors? When we design a form, often we miss some scenario we aren't familiar with. If the issue is escalated, the decision-makers can then get things moving to allow for the scenario. The goal is clear and the problem is technical. It can be fixed.
Now the rest of this is a bit different from what TFS was talking about.
However, I've often found myself in a situation where I'm asked to implement some byzantine procedure with loads of unintended consequences by what I can only imagine are committees who are stuck in the “expert system” paradigm (i.e. we can replace experience and expertise with a near-minimum wage worker in a call center navigating a decision tree, and if it doesn't work, it's only because we haven't included enough decision points). By the time it gets to my desk, the decision tree is usually completely divorced from any kind of goal or mission, and it only gets weirder as the decision points are piled on.
Some times I can work with the client and help them design something that will accomplish the mission better. The majority of clients, though, just say “Stop being rude! All you have to do is…!” (Being rude, apparently, to some people, is having the chutzpah to dare question one's betters.)
Say we need to dispatch a tow truck. Seems easy, right? Well, a call comes in from 911 dispatch and the end user marks the call as a routine matter for office hours and it gets dropped. Now the masters of the universe are (rightly) breathing down the neck of the higher-ups at the towing company. But, but! For some reason, we don't step back and see that the root cause was that the call was marked incorrectly. Now we need another decision point for 911 dispatch calling.
So, somebody calls in with minor vehicle damage after bumping into a downed tree and wondering if they should contact 911, and the operator selects 911 dispatch and sends out a tow truck. When it gets there, there's nobody to tow! Then we get an actual call from 911 dispatch, but since the truck is on the other side of the county, it takes the driver too long to get there. So, the masters of the universe get rumbling again.
Wash, rinse, repeat about 50 times, and as this decision tree gets more complicated, things get more byzantine and the frequency of these unscripted things happening increases.
At this point, I bring up the idea that perhaps the operators need a little training and we should revert our changes.
Nope, nope, can't be done! Add in more decision points! Why can't you just make the computer do the right thing?! Then I say, “Look, the reason that call was sent to the wrong tow truck was because we went through the decision tree this way and you wanted, on blue moons where the dew point is below 50 degrees and there's a 50-60% probability of precipitation and the moon is at perihelion and the CFO is at 2nd residence instead of his cottage, to send the call to that tow truck instead of the one you're saying it should have gone to. There must be some broader goal or concern behind all these decision points you've asked me to put in. Can't we set up and meeting with all the stakeholders and simplify this a bit?”
Nope, I don't have time to understand the decision tree! It's too technical! Just throw in another decision point! I'm not going to tell you why that was the wrong tow truck, only that we can't sent that tow truck if it's a 2008 Mercedes, only if it's a 2009 Mercedes, except when there's an amber alert! I'm not going to tell you the logic behind this! Just do it, you illiterate, arrogant nerd!