Despite the rise of social movements like Black Lives Matter and the greater visibility of much higher incarceration rates and arrest rates of people of color in the US, there appears to be very little movement in reforming police departments to prevent killing unarmed civilians, among other problems. The problem is bad enough in some cities that police kill civilians at rates in excess of the US homicide rate (https://mappingpoliceviolence.org/). Below I list the start of a proposal along a medical / FAA model to progressively reduce the problem.
Professionalize police forces along medical models. Police officers should have levels of training like nurses: a minimum of a 2 year college degree [+ higher pay for a 4 year degree] with heavy doses of courses in law (e.g. human rights) and human psychology; which would come in handy when dealing with many criminals, the homeless, and citizens off their medications that police so frequently encounter. Some departments currently only require a high school degree plus a 0.5 year police academy training along with a 1.5 year probation period (http://work.chron.com/long-train-cop-21366.html): essentially an apprenticeship model that understandably gives rise to wide variations in policing based on locality. And there should be no allowances for substituting military background for police education/training (unless your job was in the Military Police) - other than using a gun, the two jobs are completely different, and urban police departments with so many veterans are undoubtedly pushed into having a military occupying force mentality (plus an officer-safety-at-all-costs mentality). Finally, in the medical model, police should be required to obtain a minimum number of continuing education credits each year in order to expose them to new techniques on dealing with difficult situations and keeping up to date on the law. Yes, all this will cost taxpayer $$, but you get what you pay for.
Implement a no-fault policy for investigating police misconduct, along the model of the FAA. The no fault model of investigating aviation crashes (http://crewroom.alpa.org/alpa/DesktopModules/ViewDocument.aspx?DocumentID=3540) was absolutely critical in reducing aviation accidents from very high levels in the 1930s-1950s to the very low levels today because that allowed everyone to learn from mistakes, even terrible ones, without the specter of "accountability" hanging over everyone involved. Similarly, investigations of police misconduct, including killing unarmed civilians, should provide immunity to the police officers involved with only 2 exceptions: a) the police officer killed the civilian in a premeditative, illegal fashion (i.e. a dirty cop who committed a crime on purpose), or b) the officer is found to have concealed any information associated with the incident. The penalty for a) is prosecution, and the penalty for b) can be loss of a job, pay, etc. Thus, in this model there would be no need for the thin blue line: cops covering up for each other, because now they all have immunity from consequences provided they are completely truthful (same goes for investigating officers, coroners, etc). Of course, such a model should include automatic $$ payouts to families of innocent victims of police killings, e.g., to indicate that their harmed family member was valuable to the community (w/ incredibly high $$ additional payouts if police attempt a coverup [e.g. planting a gun on a victim]) in addition to families being more likely to find out the truth of what happened, which itself is very important.
Independent state bureaus collecting all accident data plus outside accreditation councils that can force protocol changes on police departments. In exchange for no-fault police investigations, there must be very strong mechanisms in place to force changes on police procedures to reliably, incrementally make policing safer for citizens. Police department are required (if they want to keep their funding levels) to report all investigations to a state body that will then routinely make binding changes to police procedures/training/education in the state or for a particular department. Also, all police departments (like medical institutions must) must undergo accreditation every few years under the purview of an independent national body that verifies that required procedures are being followed, that personnel have proper training, and that citizen safety and rights are at adequate levels. If not then the auditing body can recommend the department be put into state or department of justice receivership until the problems are corrected: no waiting around for an ACLU lawsuit to force the issue when abuses reach high levels.
At first it does seem counter-intuitive to try to address abusive/racist policing by lessening personal responsibility of the officers committing the injustices. However, one has to admit that our current system of accountability isn't working: no matter how egregious the unjust police action is, most actions are found to be within procedures, or if not prosecutors are reluctant to prosecute, or if under political pressure when they prosecute the prosecutors don't try to convict very hard (they undercharge or overcharge the officer, e.g.), and finally juries are reluctant to convict. Thus, what have we got to lose (besides more $) by trying another model (successful in other fields) of reducing police abuse/accidents?