The 4th Annual Meeting of the Neurophilosophy of Free Will Consortium
June 27th~30th, 2023
The Sigtuna Foundation, Sweden
(All meeting sessions except for the public lecture are only open internally to the consortium members. Any
dissemination or distribution of the private meeting sessions is strictly prohibited)
From Airport Stockholm Arlanda: From Arlanda 20 minutes by taxi. Taxi Stockholm. Phone +46 8 15 00 00 or https://www.taxistockholm.se/. By public transportation, bus SL 583 to Märsta Station. At Märsta station change to bus SL 575 (get off at bus stop Sigtunastiftelsen). For timetable, see sl.se. Tickets can be purchased on the bus with contactless bank cards, in the SL-app, at Metro and commuter rail stations and at ticket agents such as Pressbyrån and 7-Eleven.
From Stockholm with public transportation: With SL commuter train sl.se to Märsta or SJ train sj.se to Märsta. At Märsta station change to bus SL 575 (get off at bus stop Sigtunastiftelsen). For timetable, see sl.se. Tickets can be purchased with contactless bank cards, in the SL-app, at Metro and commuter rail stations and at ticket agents such as Pressbyrån and 7-Eleven.
By car from Stockholm:Highway E4 north heading towards Uppsala, turn at exit sign: Sigtuna, Märsta, Arlandastad. Follow the road sign posted Sigtuna road 263. At Sigtuna’s roundabouts: follow the signs “Hotellslinga S” Hotellslinga S, that leads you to Sigtunastiftelsen.
Link with direction on Google maps
Consciousness and the Fetzer Institute
Update on ongoing studies at the Brain Institute and discussion of future directions in the field
Uri Maoz & Aaron Schurger
We will give an update about the state of ongoing studies at the Brain Institute that are part of the Neurophilosophy of Free Will project. We will also begin a discussion about future directions in the field of the neuroscience of volition, to initiate a longer discussion that we will have on Friday morning.
How does the brain stop us from making mistakes?
Imagine writing an email when you are upset. As you reach toward your phone to click “send”, you realize sending it would be a mistake; you change your mind and stop yourself from clicking. Once we act, how can we change our mind? “Free will” is a topic of lively philosophical inquiry but has limited grounding in physical observations. In the first part of this talk, I will show results from a change-of-mind (CoM) paradigm in which rats mistakenly act and then decide to stop the in-progress action. In the task, head-fixed rats discriminate two stimuli by either running on a treadmill past a distance threshold (Go) or remaining immobile (NoGo). On CoM trials, rats mistakenly begin running to the NoGo stimulus but soon realize their mistake and chose to return to immobility before crossing the threshold. We found two anterior cingulate cortex (ACC) neuronal ensembles for monitoring and adjusting actions that are distinguished by their operational timeframe. One performs real-time action monitoring to enact CoM, while the other looks back in time to monitor recent actions and outcomes. In the second part of the talk, I will discuss the neural causes of mistaken actions. Distinct groups of ACC neurons selectively respond, either to the Go, or the NoGo stimulus. On CoM trials, Go stimulus-preferring units aberrantly responded to the NoGo stimulus. If this stimulus-evoked response was larger, then the rat mistakenly ran faster. Thus, ignition of the wrong pool of stimulus-responsive neurons may drive mistaken actions. Finally, I will show that motor cortex EEG beta oscillations cause action stopping during a CoM and highlight their utility for controlling brain-machine interfaces.
The role of prestimulus and stimulus-elicited neural activity in eliciting conscious visual perception and moral responsibility
Prominent theoretical views have debated the role of consciousness in eliciting moral responsibility. One view holds that although an agent can be responsible for actions and outcomes that are affected by unconscious perceptions, they could be even more responsible if they were conscious of the relevant information. Empirical evidence has likewise shown that participants assigned higher responsibility to agents who consciously perceived morally relevant visual information compared to those who unconsciously perceived it. One challenge with this view is that people’s ability to perceive relevant information consciously and accurately is influenced by the neurocognitive state of the brain prior to the appearance of relevant visual information. In such cases, people do not have control over whether their prestimulus neurocognitive brain state will facilitate, impair, or distort accurate conscious perception. To explore some of these issues, we will present recent empirical work that evaluates (a) the cascading effect of prestimulus and stimulus-elicited neural activity on conscious and/or accurate visual perception and (b) how lay first-person responsibility judgments varied according to this conscious and/or accurate visual perception. We will also discuss how this work can inform competing philosophical theories about consciousness, control, and moral responsibility. Last, we will outline planned research that could extend our conclusions to different kinds of reason-responsive actions.
The sense of agency in a context of reduced freedom
The sense of agency is frequently considered as being tightly intertwined with freedom of choice, with a greater sense of agency being associated with a greater number of alternative actions and decisions. However, freedom of choice is far from being guaranteed in many contexts and situations. In this talk, I will first review studies showing that when people have a restricted number of options and rather follow the orders of an authority, their sense of agency is reduced compared to a situation where they are free to decide. I will then review several studies conducted with populations facing a daily restriction of freedom. Many social structures indeed rely on strict hierarchical organization where people are required to follow rules and orders. I will address how evolving how highly hierarchical and sometimes coercive structures restricting human’s autonomy impacts the sense of agency, even when one is free to choose which action to perform. More precisely, I will present several studies conducted with the military and with inmates in prisons with different detention regimes. Overall, we observed that such contexts restricting autonomy have a negative influence on the sense of agency, reducing it when people can make free decisions. In inmates, we even observed that they actually have a greater agency when they follow orders compared to acting freely, probably as a result of social isolation. We will also review how being trained to be accountable for one’s own action influence the sense of agency.
The six pillars of volition: a conceptual framework for the scientific study of human free will
Jake Gavenas & Uri Maoz
Volition is fundamental to goal-directed behavior and separates humans and other complex animals from simpler organisms. However, the lack of an agreed-upon definition of volition has complicated the scientific study of volition and led to conflicting results surrounding the neural mechanisms of “volitional action”. We will introduce a conceptual model of volition, defined as an umbrella term for a set of interrelated behavioral capacities (akin to “cognition”) rather than as a singular phenomenon. We first outline some of the common dichotomies employed in the study of volition, such as comparing free choice to instructed actions or acting spontaneously to acting in response to an external signal. From there, we introduce six behavioral capacities that act as “pillars” of volition: endogenous action initiation, decision-making, self-control, option generation, intention maintenance, and agency. We outline recent research that falls into each of these six areas, compare our model against prior frameworks for volition, and describe the ways in which conscious awareness interfaces with each pillar. Our intention is to introduce the model briefly and leave ample time for discussion.
Epistemic agency, intentional action & extended volition
In this talk I want to present one half of the core argument of the Tinkering Mind which holds that if epistemic agency is both a form of intentional action and a form of cognition then extended cognition has to be true. I then want to apply the argumentative strategy of this argument to the case of volition.
The core argument builds on Yair Levy’s work which shows that there is no functionally relevant difference between intentional epistemic actions inside the head and environment involving epistemic actions. I argue that this argument succeeds and that because of it epistemic intentional environment involving actions are literally a form of cognition and extended cognition is true if we, very plausibly, hold that intentional epistemic actions inside the head can be constitutive parts of cognition.
In the second half of the talk, I apply this argument to the case of volition. Volition can mean many things. Here it is understood as active choice. Choices which are our actions rather than something that happens to us. I want to argue that if such active choices exist then they are subject to the same kind of argument as epistemic intentional actions. This means that such choices do not just happen inside the head but will often contain full-fledged environment involving actions. In other words, such active choices allow for extended volition.
Modulating the Readiness Potential via Theta-Burst Transcranial Magnetic Stimulation
However, an accumulation-to-bound model suggests that RP onset may not indicate motor preparation (Schurger et al., 2012). This model further provides a testable hypothesis that pre/SMA inhibition would modulate response times in arbitrary—but not deliberate—decisions and decrease the early components of the RP (Maoz et al., 2019). Here we propose to test this hypothesis using theta-burst stimulation (TBS), which is a non-invasive transcranial magnetic stimulation protocol that creates safe, reversible inhibition of the human cortex lasting for about half an hour. In particular, we inhibit M1 and pre/SMA activity (in separate trials) in an attempt to diminish the RP and affect behavior as above.
Autonomous behaviour and the limits of human volition
The nature of volition and autonomy is one of the longest-running scientific debates. Philosophical accounts of free action distinguish between freedom from external constraint and freedom to achieve personal goals and desires. Many studies of volition focussed on the “freedom from” aspect of voluntary action by creating conditions where action is or is not independent of an external triggering stimulus. However, such studies neglect the reasons or goals that an agent may have for acting, and omit the crucial “why” aspect of voluntary action.
Here we propose that some competitive game scenarios can capture both the “freedom from” and “freedom to” aspects of voluntary action. For example, in the ‘rock, paper, scissors’ game, each player selects an action without first seeing the action of their opponent, while at the same time, innovating and discovering a specific action that defeats their opponent. In this study, we examined how free and how flexible such reasons-guided voluntary actions can be by identifying whether people can become free from internal constraints on behaviour in adaptation to competitive pressures.
Voluntary actions and problem-solving
Volition refers to a capacity to initiate actions through one’s autonomous conscious decisions. Voluntary actions are often made for a reason, which is typically a representation of a desirable goal state. However, existing paradigms for studying volition focus on arbitrary, meaningless actions, favouring the stimulus independency aspect of volition rather than goal-directedness (Haggard, 2019). The aim of this study was to develop a new view of volition as goal-directedness. We investigated the neurocognitive processes of goal-directed actions, their overlap with the brain areas/processes traditionally associated with voluntary actions, and their interaction with other neurocognitive systems during a problem-solving task.
Semicompatibilism: Integrating views on free will and responsibility?
John-Dylan Haynes & Walter Sinnott-Armstrong
The implications of psychological and neuroscientific findings for the problem of free will have long been debated. On one side of this debate neuroscientists have often considered choice-predictive brain signals in simple decision tasks as evidence for determinism and thus also as challenges to responsibility. On another side of the debate the representativeness of the typical experimental tasks for free choice has often been challenged. One typical view has been that only decisions involving some form of deliberation and reasoning can be considered free, thus rendering the neurocognitive studies irrelevant because they involve seemingly random choices. This view renders determinism and free choice compatible and thus also provides a base for responsibility. While this theoretical conceptualization is tenable, we have shown in several studies that it collides with lay views on the nature of free decisions. Laypeople tend to consider exactly the seemingly random, unconstrained choices without consequences to be free, thus the opposite of the reason-based view. Here we instead follow the different path that can salvage responsibility in a determinist universe (without requiring any decision about whether this determinism is true) while at the same time maintaining compatibility with lay views. The path is to completely dissociate responsibility and free will, which might appear counter intuitive, but is again in line with lay positions. One version of this is so-called “semi-compatibilsm” as advocated by John Martin Fischer. Here, we will give a joint perspective on semi-compatibilism from the perspectives of psychology and neuroscience on the one hand and philosophy on the other.
On the origin of intentions
Consciousness presumably evolve through interaction with the environment, eventually including an awareness of intentions and a sense of a free will. Intentionality, as goal-drectedness, would include the creation and projection by the brain of alternative future states, desired, avoided or feared. Together with perception, intentionality can be regarded as a key aspect of (conscious) behavior in perhaps all animals. In the Agora research group, we are interested in the origin and evolution of intentional processes that may lead to decisions and actions. Hence, we have developed neurocomputational models of various brain areas believed to be involved in these processes. The aim is to use model simulations to complement experimental studies in exploring causal pathways of signals in the neural process of volition, and to test hypotheses by suggesting new experiments. In this session, we will present and discuss two complementary models, addressing the questions “when”/”whether” and “what” to act, respectively. The former is a more abstract model of the cortical-thalamic-basal ganglia loop and the latter is a more detailed attractor model of the network dynamics of LPFC and ACC, and their interactions.
Proximal intention and possible subcortical generators of the readiness potential
Intending an impending action includes a combination of excitatory and inhibitory processes for response timing or choice selection. While these may be initiated from the prefrontal cortex, subcortical areas also play a key role. Indeed, the readiness potential (RP) may be a neural correlate for this. Subjects that have motor and cognitive disorders such as OCD may exhibit stronger RPs while subjects with Parkinson’s disease may show weaker RPs. In both of these disorders, the basal ganglia and dopamine are factors. A dynamical systems model was developed to explore brain area dynamics during simulated tasks for self-paced and instructed action. It shows how inhibition may dominate during the variable-length early RP and execution may dominate during the more fixed-length late RP and LRP. The modeling dynamics also show how dopamine levels may effect reaction time and RP amplitude.
Modeling the neurodynamics of intentional processes using attractor networks
Azadeh Hassannejad Nazir
The brain can be considered a nonlinear system, which displays complex neurodynamics, including both ordered and disordered activity patterns. This prominent feature can provide a basis for adaptation and goal-directedness of actions. Fundamental questions related to cortical network dynamics is whether intentional decision-making relies on transitions between different activity patterns, and whether any specific type of neurodynamics could be associated with consciousness. To investigate these questions, we use attractor neural network models of two cortical structures involved in volition, LPFC and ACC. These network models and their interaction can display different dynamics thought to correlate with deliberate action control during intentional processes. Different neural activity patterns represent different functions, such as preparatory processes of intention, implying a volitional decision-making process. In our model, the interactions between neural attractors, correlating with different optional choices, address the “what to act” aspect of volitional control. This aspect is characterized by time, frequency and phase of the oscillatory EEG-like simulations.
How is deliberation related to volition? An EEG-based paradigm
Historically, the neuroscience of volition largely focused on arbitrary (meaningless) decisions and sought to generalize findings to deliberate (meaningful) decisions. However, more recently there have been claims that the neural correlates of arbitrary and deliberate decisions might be different enough to challenge such generalizations. In particular, some studies suggested that the readiness potential (RP) is at least greatly diminished in deliberate decisions. However, other studies did not find differences between the RPs in arbitrary and deliberate decisions, leaving the matter unresolved. We therefore decided to again compare arbitrary and deliberate decisions, but this time using four different measures related to volition—the RP, the sense of agency (measured using the temporal-binding effect), willpower (measured with handgrip dynamometers as the effort exerted to indicate the decisions, using the concept of willpower as a limited resource of volition that requires effort) and the subjective sense of volition—as this would enable us to compare these two types of decisions more widely. We recruited 50 participants, who were instructed to decide between two options to maximize monetary reward. The two options could vary along two dimensions with different levels of values that were used to fill gauges. When filled, those gauges increased a monetary reward (from 15€ up to 22€) in deliberate conditions, but not in the arbitrary condition (where the gauges were associated with a 0€ increase). In the easy deliberate condition, only one dimension of the two options had to be compared to select the most rewarding option, while in the hard deliberate decision both dimensions needed to be compared, thus requiring more deliberation. We found evidence that there are no differences among the 3 conditions for the RP and sense of agency. However, we observed both a higher willpower and higher reports of subjective sense of volition in both deliberate conditions compared to the arbitrary condition. Moreover, we found evidence for no differences between the two deliberate conditions regarding willpower and the subjective sense of volition. These results indicate that there is a qualitative contrast between arbitrary and deliberate decisions concerning the effects on willpower and the subjective sense of volition, as quantitative differences in deliberation did not influence these measures. This study thus suggests that there are likely neural differences between arbitrary and deliberate decisions—although, in this task at least, those were not expressed in the RP. Our findings therefore further challenge the generalization of results from arbitrary decisions to deliberate decisions when studying volition.
Volition, Introspection, and Imagination
Let us define “decision” as the mental event that settles the question what to do. Let us define
“conscious” as knowable directly, not through interpretation. Are there conscious decisions?
Some have argued that no: we always come to know our own decisions indirectly, after the
action. They have also argued that there is a neural event preceding the action — of which we
have no knowledge at the time — that corresponds to the decision. Others have argued that
yes: at least sometimes, we come to know our own decisions directly, before the action. They
have argued either that the neural event preceding action happens at the same time as we
come to know the decision, or that it does not correspond to a decision after all.
Evidence from accumulator models for early vs late decisions
A friendly debate between John-Dylan Haynes and Aaron Schurger
Studies on the readiness potential and other choice-preceding brain signals have long played a major role in the debate on free will. Such signals are often believed to suggest that the outcome of “free” decisions is determined prior to the time of the subjective decision. Recently there have been several challenges to this interpretation. Computational modelling work has suggested that the readiness potential might constitute a mere artefact of backwards averaging of stochastic brain signals, with the actual decision point occurring later when a decision threshold is crossed. In another line of work, real-time neurofeedback has suggested that an emerging decision can still be aborted, up to about 200 ms before movement onset, thus also questioning the causal role of the readiness potential. However, several questions remain open. Here we will debate (a) the evidence for accumulator model accounts of the readiness potential and (b) the evidence for or against early versus late decisions. We will then work to jointly develop an adversarial framework for testing its predictions.
Volition in non-human animals and machines
Our discussions have largely centered on volition in humans. In this talk, I would like to move away from this comfort zone to brainstorm about what would constitute evidence for or against volition in non-human animals and machines. First, I will provide an update on collaborative work that we have conducted in the C. elegans worm, showing that we can create a hybrid between artificial and biological neural networks. This hybrid can control the volitional decisions of the worm and maneuver the animal towards specific targets, food, generalizing over obstacles and across individuals. Second, I will shift gears completely and show results about Turing tests designed to evaluate whether language answers come from humans and machines. I suspect that most of the audience will be scandalized by notions of volition in non-human animals and machines. Therefore, at the end, I will conclude with some provocative remarks to open up the discussion about what would or would not constitute evidence that an agent shows behavior shows volitional behaviors.
Project summary & future directions
Do unconscious influences on decisions threaten free will?
Adina Roskies & Walter Sinnott-Armstrong
An agent’s awareness of reasons seems essential to flexible and rational decision-making, at least if one cannot respond to reasons of which one is unaware. However, social psychology has shown that many factors, including implicit biases, can affect decisions through non-rational and non-conscious means. We will discuss whether these unconscious influences on decision and action undermine free will and moral responsibility, and if so, under what conditions.