This post is a continuation of a series of posts about how government by data obligates us to follow the recommendations of analytics from data. Big data based analytics is an evidence based decision making. Many people see evidenced based decision making as ideal or at least believe that decisions should not be contrary to evidence.
As I discussed here, the ideals of evidence-based decision-making can obligate the decision maker to follow the recommendations from data. This kind of obligation erodes the accountability of the decision maker and thus the opportunity of the affected population (perhaps injured by the decision) to address their grievances because the decision maker’s only defense to the decision is that the data gave him no choice and that in the future he has no choice but to continue to follow data decisions. With automated big-data analytics, we have the opportunity to simply eliminate the human decision maker and many instances (such as in marketing campaigns) we have already done that.
In another post, I asserted that human accountability is essential for maintaining social order because it provides the opportunity to address grievances that permits the building of a super-majority of consent to be governed by a lesser majority. When we burden a human with the responsibility to defend his earlier decisions, that decision maker will demand to be personally convinced of the merits of the decision. Especially in the era of big data (of huge volume, velocity, and variety) the recommendations from big data analytics are incomprehensible to the decision maker. He may understand the concepts of the algorithms, but he will never be able to independently verify the recommendations by studying the data himself. If we insist on holding him responsible for defending any decision, we need to allow him to satisfy his doubts about the analytic results and that leaves the possibility of making a decision that contradicts the evidence because the evidence doesn’t overcome his doubts.
The goal of accountable leadership conflicts with the goal of obligating decisions based on evidence such as recommendations obtained from big data analytics. If we obligate the decision maker to make a decision, he can not be personally accountable for the decision. If no human is accountable for a decision, then we risk losing super-majority consent necessary to defend the government from dissenters.
Big data analytics increasingly obligates decision making. The resulting loss of accountability risks social disorder. I wrote earlier that one answer to preserve social order is to obligate the population’s participation to cooperate with these decisions. Evidence based decision making obligates both the decision making and the cooperative participation of the affected population.
In that post, I referenced an article about doctors choosing to decline medical treatment for their potentially fatal illnesses because of their experiences of watching their patients go through similar treatments with little or no success. Even though treatments are available for their conditions, and even when those treatments offer better odds of survival, some doctors decline the recommended treatments. They make a rational choice that they have a better chance of enjoying their remaining days if they stay out of the medical system.
I agree that this is a rational decision. Initiating a treatment process immediately distracts a person from enjoying his life. Often the treatment has an immediate impact on degrading the quality of life through painful procedures that can be painful or at least result in significant restraints on allowed activities. Even if the treatment starts mildly, there is the inconvenience and mental anxiety of arranging appointments and waiting for test-results that dictate very different future treatments. For illnesses with a high potential of being fatal, the treatments will compound the suffering perhaps with the same fatal outcome.
One rational decision is to decline the recommendation. But declining a recommendation is to make a decision contrary to the evidence. For example, a recommended treatment may result in significant improvement of survival probabilities. Perhaps an illness untreated will result in 5% chance of survival while a treatment may improve that chanced to 15%, a three times improvement based on high confidence data. There remains a 85% chance the patient will still die, but the evidence-based decision is that treatment is superior to non-treatment.
The above mentioned doctors have the choice to decline treatment as do all patients. However, the concept of evidence-based decision-making obligates everyone to follow the recommended treatment. Both the decision maker and the affected population (in this case it may the same person) needs to obey the recommendation.
In the context of the modern health-care objective that relies heavily on data to manage costs and optimize care, we need to coerce patients and their doctors to follow the treatment (or non-treatment) recommendations from the data analytics. In order to improve the performance of the data analytics we need new data about what happens when the recommendations are followed. Consistent data requires following these recommendations diligently and without hesitation. The implementation of the recommended procedures provides the necessary data to help refine the algorithms to better decide what patients are more likely to benefit from the procedure.
The imagined promise of a more perfect health care system needs more data informing us what happens when a recommendation is followed. In order to obtain that data, the recommendations must be followed consistently. In a sense, the obligation to participate is the requirement to sacrifice one’s own suffering through a potentially unhelpful procedure so that more successful recommendations may be available for future patients.
Evidence-based decision making imposes an obligation to participate even when it becomes an obligation to suffer.
The above summary of my earlier criticism of aggressive imposition of following big-data analytics is background for my reaction to a recent news of a young lady who has chosen to end her life with dignity when facing a terminal illness involving an aggressively growing brain tumor. She presents her case here. After reading her story, I tend to support her decision (assuming she follows through). I hope I would have a similar opportunity if presented with a similar prospect.
In contrast to my earlier example of experienced doctors diagnosed with terminal illnesses declining treatment to all the disease to take its natural course, this example takes an additional step of ending life while life is still tolerable. I accept that the doctors were rational to decide their remaining days will be better enjoyed by avoiding treatment. If the goal is to enjoy remainder of life, then it should be rational to end the life before that life becomes impossible to enjoy. In the case of Brittany Maynard’s decision to set a specific date, she will be ending life many days (perhaps months) before life becomes impossible to enjoy. She is making the decision to end life while she can still can enjoy life, and to avoid as much of the inevitable suffering as possible.
It is brave of her to make this choice. I am further impressed by her efforts to use her example to raise awareness of the opportunity for the terminally ill person to choose a date of death before when death would occur if the disease takes its natural course. This publicity has reminded us of the old debate about the ethics of suicide especially in terms of facing terminal illness. Quoting an NBCNews article:
Brittany is 29, newly wed, attractive, articulate and terminally ill. She has a highly aggressive, fast-growing brain tumor that will probably end her life in less than six months. She has suddenly become the new, self-proclaimed face for the right to die.
Despite the inevitability of terminal illnesses to suffer at a minimum the lost opportunity to continue to enjoy life, the public sentiment still disapproves of the option for suicide to prematurely end the life. Only a handful of states in USA allow physicians to prescribe a lethal dose, and even these have strict procedures about when this is permitted. There are numerous arguments against offering patients the option of suicide. Some of these are summarized in this earlier article at USNews and this specific response by another brain-tumor patient who argues against suicide. To me, one of the stronger arguments against permitting doctor-assisted suicides is the problem of coercion as discussed in the USNews article. Patients may feel pressured to end their lives early to avoid being a burden on the family, disposing of savings that would otherwise be inherited, or being a burden to the health care and insurance industries as a whole. It is hard for us to be convinced that the decision to end life does not involve such coercion.
I have read some criticism of Brittany Maynard publicizing her decision in advance, but I think this publicity is justified by the fact that it provides a convincing case that she is making this decision for the understandable reasons. This allows her to make the case that she is not making this decision to avoid being a burden on her family or on society. Instead, she is making the decision based no her own wishes to die while she still has dignity of a sound mind and a life that she can enjoy. As noted with the above response, her appeal will not satisfy many people.
Her case is not going to resolve the debate, but at least is presents another example of someone who could have many more days or even months of a life where she can contribute something to her family and community. She is providing an example to allow us to continue this debate and as the above NBCNews article states:
Still, a whole new generation is now looking at Brittany and wondering why their state does not permit physicians to prescribe lethal doses of drugs to the dying. Brittany is having and will have a big impact on the movement to get measures before voters or legislators.
I think this is a debate the deserves more attention. My feeling is that such life-ending options should be available in all states so all terminally ill patients have the option of dying before the disease robs them of their dignity.
For this post however, I argue again my preference by presenting a different argument. Following the above concepts of strict obedience to evidence-based decision-making, we can not permit people to decline an evidence-based decision to proceed with treatment that has the potential to extend her life even if the resulting life is degraded through handicaps and disfigurement. The decision must be followed if there is persuasive evidence that the treatment offers better survival odds than no treatment. If we accept the supremacy of evidence-based decision-making, then we must demand this decision and the obedient following of this decision.
From the perspective of big-data analytics (or data science), in order to make good predictions we need good data. In the area of medicine, we generally need much more data than we already have. This is especially true for rare conditions like terminal brain tumors. In order to determine the effectiveness of the recommended procedure, we need to collect data for what happens when the evidence recommends the procedure. Even for common conditions, there is such a wide variety of health-relevant variables in the entire population of individuals that we need abundant measurements in order to separate those who benefit the most from those who benefit the least. In order to have data to identify those who benefit the least from the recommended procedure, we need to observe this outcome from their following the recommended procedure. In order to identify the bad recommendations we need to observe them resulting in bad outcomes.
Government by data implies an obligation to follow the recommendation. Decision making based on big-data in the context of medicine obligates us to suffer through the struggles between medical science and nature to determine better medicine for the future. The present patients need to accept the recommended procedures in order for us to measure the effectiveness these recommendations. This information will allow the predictive analytics to make better decisions in the future.
In the earlier post on obligation to participate, I argued that big-data analytics should not allow patients to decline recommended procedures. Here, I argue that it should not allow patients to take their own lives. The supremacy of evidence-based decision-making demands the collection of new evidence. A patient must suffer through to the natural end of life resulting from recommended treatments.
This obligation is part of a web of preconditions that will permit big-data to produce ultimately beneficial predictions. The abhorrent conclusion of denying a person a dignified choice for dying is a necessary consequence of earlier decisions to automate decision-making (decision-makers can not make decisions contrary to the evidence) and to preserve social harmony (participates are obligated to cooperate, because there is no accountable human decision-maker). To allow big-data to save health-care, we have to let big data make the decisions instead of doctors or patients making subjective decisions of what they think would be best in their present circumstances.
Even when the physician-assisted suicide is available, it is pretty rare for people to have a certain diagnosis of terminal illness to qualify. Many more people will die suddenly or unexpectedly for conditions that are not considered terminal illnesses. They will suffer without the opportunity to choose suicide earlier. Many people will die horribly in traumas and other deaths that had no opportunity to end life before the bad happens. While it makes sense to give someone with a terminal illness diagnosis the opportunity to die early instead of following recommended health or hospice care, the number of people who will qualify for this option is probably a small minority of all deaths. I suspect most people will not benefit from this option. Meanwhile, denying this opportunity allows us to observe nature taking its course for the specific combinations of traits for each patient. This provides us data that can be useful in the future to make even better recommendations. From a data science perspective, we should not permit people to deny recommended treatment or to escape the natural progression of an illness.
From my personal perspective, I hope the options to decline care and even accept assisted suicide will be available when my time comes. I regret that there may be good reasons why I may not have that option.