Neuroscience, Free Will & Moral Responsibility

Has neuroscience proved that free will is an illusion?

Over the past few decades, research in neuroscience has suggested that upcoming random decisions (such as raising the left or right hand without reason, purpose or consequences) could possibly be decoded from the brain before humans become aware of having decided. This is due to work that showed that information about upcoming action exists in the brain before human subjects report having decided which action to carry out and when to act (e.g., Libet et al., 1983Haggard & Eimer, 1999; Soon et al., 2008Fried et al., 2011).

It was therefore argued that consciousness might not play a causal role in what are experienced as free decisions. This, if true, has profound implications for the debate over free will, and could challenge some pillars of our social order, like our notions of moral and criminal responsibility.

However, these far-reaching conclusions hinge on positive answers to all of the following four questions. 
  1. Are the neural signals, that are correlated with the decisions and precede their reported onset, actually associated with the decision process? 
  2. Can subjects accurately report the temporal onset of their decisions? 
  3. Decisions that we sincerely care about, that showcase our free will, are deliberate: reasoned, purposeful, and bearing consequences. Can the conclusions from experiments that found pre-conscious neural information about upcoming random actions be generalized from random to deliberate decisions? 
  4. Can the results of such “retrodiction” experiments, where the neural data is analyzed after the fact, be replicated when attempting to predict the decision – i.e., when the analyses are carried out on the fly, online and in real time? 
    • This is important because various technical difficulties were found with such retrodiction analyses, arguably necessitating online, real-time analyses to convince of the existence of predictive neural information about upcoming deliberate decisions.

Over the past few years, these questions have become a main focus on my research. 

1. Predeliberation neural signals are predictive of action

To begin answering the first question, I initiated a collaboration with monkey neurophysiologist and decision-making expert Prof. Daeyeol Lee of Yale University and Prof. Ueli Rutishauser of Cedars Sinai Medical Center. This has led us to the discovery of bias neurons. These neurons especially influence random decisions even before the decision alternatives are known, and thus before rational deliberation begins, in the monkey prefrontal cortex and basal ganglia. We found both anatomical and functional dissociations between these neurons and neurons that code the decisions of the animal during the deliberation period (Maoz et al., 2013). As this bias activity was a novel computational mechanism, so Prof. Rutishauser and I developed a circuit model that reproduced our results and made additional predictions, which were borne out empirically (Maoz et al., 2013)

Conclusion: The neural signals captured in the random-decision experiments above may have more to do with predeliberation bias activity than with the actual decision process.

2. Humans have systematic, large errors in reporting decision onset

As for the second question, there is increasing evidence, including from my collaborative work with Dr. Liad Mudrik, Prof. Gideon Yaffe and Prof. Ram Rivlin, that the reported timing of decision onset is systematically biased and inaccurate. It is not even clear that the onsets of our decisions are consciously available to us (if such decision onsets are even distinguishable neural events; Maoz et al., forthcoming). 

Conclusion: It may be best to stop using the timing of the subjective reporting of the decision onset as a measure for actual decision onset.

3. Deliberate & random decisions appear to rely on divergent neural processes

Investigating the third question, my research with Dr. Liad Mudrik shows that the early neural signals correlated with the selected action for random decisions do not replicate for deliberate decisions. Instead, such correlated signals appear to be finalized only very close to movement onset for deliberate decisions (Mudrik, Maoz, et al., In preparation)

Conclusion: The results identifying early neural signals correlated with the decision in random decisions do not seem to generalize to deliberate decisions.

4. Early, predictive signals so far not found for deliberate decisions

To make progress with the forth question, I have been collaborating with neurosurgeon Prof. Adam Mamelak and neurologist Prof. Jeffrey Chung at Cedars Sinai Medical Center and with neurosurgeon Prof. Ian Ross of Huntington Memorial Hospital. They granted me rare and unique access to epilepsy patients who were implanted with intracortical electrodes as part of their clinical presurgical evaluation. I was therefore able to record local-field potentials and single-neuron activity directly from the human brain. I have now had the good fortune to have worked with about 20 such patients on various tasks related to decision making and consciousness. This has, among other things, enabled me to develop a first-of-its-kind online, real-time system for action prediction in a competitive environment. Using this system, we were able to show, for the first time, that action contents can be decoded from brain activity before action onset online and in real time (Maoz et al., 2012). Additional results, focusing on the neural dynamics of decision processes, are now being prepared for publication (Maoz et al., in preparation). But, so far, neither we nor others have been able to demonstrate the predictability of action contents before the onset of deliberate decisions online and in real time. We think that this is because such deliberate decisions are not strongly influenced by the bias activity we measured in the monkeys above, on the one hand, and is finalized only very close to the moment subjects become aware of them, as discussed above, on the other hand. 

Conclusion: The early predictive neural signals characteristic of random decisions have so far not been found for deliberate decisions.

Overall conclusion: Neuroscience has not proved free will to be an illusion

Given the state of the overall empirical evidence gathered so far, drawing strong conclusions against a role for consciousness in decision making appear too hasty.