Computational Theories of Mindfulness

By Liza Tatishev

Mindfulness involves non-judgmental moment-to-moment self-awareness of physical and mental states– it is the ability to be fully conscious and aware. Within the past decade, scientific interest in mindfulness practices has increased. Mindfulness has a long history: originating over 2,500 years ago in Indian Buddhist tradition [1]. By the mid-twentieth century, mindfulness became secularized and incorporated into psychological intervention methods to improve emotional well-being [2]. Today, mindfulness-based intervention methods are recognized as successful approaches in treating a variety of psychological disorders, including depression, anxiety, and addiction.

A more precise understanding of mindfulness can benefit the refinement of mindfulness-based intervention methods. The field of computational neuroscience offers a method to examine the neuroscience of mindfulness. In order to approach the neuroscience of mindfulness computationally, it is important to first clarify the physical reality of the brain that is being represented via algorithms.

Psychological Mechanisms

As a psychological intervention method, mindfulness practices– such as Tibetan Buddhist imagery meditation [3], Zen meditation [4], and Kundalini yoga [5]– are grounded in training and refining attention skills. Attentional control is a key cognitive process of executive functioning (also referred to as cognitive control) necessary for emotional and behavioral regulation [6].

Cognitive control

Mindfulness training has been shown to enhance cognitive control in children and adults [7]. On a phenomenological level, the process of mindfulness training engages with the self-regulation of cognitive states that are interrelated with attentional networks.

Within cognitive psychology, attention consists of sub-processes (including alerting, sustained attention, and conflict monitoring) that underlie the ability to attend to stimuli. Higher self-reported mindfulness is related to higher measures of more stable sustained attention [8]. Furthermore, longitudinal studies have found that engaging in mindfulness practices correlated with improvement in sustained attention tasks over the course of intervention [9].

Sustained attention recruits attentional control functions, while inhibiting proponent, but non-relevant responses [10]. Research has demonstrated that mindfulness may improve performance on attentional control tasks, however experimental designs should be improved to remove discrepancies in the field [11].

Evidence for mindfulness practice improving sustained attention and attentional control processes exists. Enhanced cognitive control network activity is associated with increased psychological well-being [12] [13] [14].

Therapies

Mindfulness offers a framework of self-awareness, -regulation, and -transcendence that can be implemented in psychological intervention methods [15]. Mindfulness is now incorporated as a major component of empirically supported psychological intervention methods, including mindfulness-based stress reduction (MSBR) [16], mindfulness-based cognitive therapy (MBCT) [17], acceptance and commitment therapy, and dialectical behavior therapy [18].

Neurophysiological mechanisms

The cognitive and behavioral effects of mindfulness can be examined through the theoretical framework of emotion regulation. Studies on the neuroscience of mindfulness demonstrate that individuals who engage in mindfulness demonstrate functional and structural differences in the emotional regulation network– specifically, by strengthening the ability to regulate ruminative thinking [19]. Effective emotional regulation facilitates psychological well-being [20].

Brain activity

Early electroencephalogram (EEG) studies and behavioral studies have demonstrated that mindfulness practices engage particular attentional networks. Mindfulness practices produce state changes that increase activation of core nodes in the dorsolateral prefrontal cortex (DLPFC) of the central executive network (CEN) (areas associated with executive function and attention) [21]. Other associated neural networks that show increased activity with mindfulness practices include the anterior subdivision of the cingulate cortex (ACC) [22] and the insular cortex (IC) [23].

Mindfulness training can result in the upregulation of underlying neural networks in brain activity at a resting state [24], which is correlated with significant improvement in emotion regulation [25]. Causal evidence for neurophysiological mechanisms underlying the cognitive benefits of mindfulness training exist. For example, children who receive mindfulness training intervention exhibit a stronger preserved resting state default mode network (DMN)-DLPFC anticorrelation (decreased activation of the DMN and increased activation in the CEN) compared to controls [26]. Enhanced DMN-DLPFC anticorrelation correlates with better sustained attention performance [27].

Computational theories

Attentional control and emotion regulation are prime targets for computational modeling. However, comprehensive computational frameworks on the effect mindfulness practises have on emotion and cognition are nascent [28].

Bayesian brain hypothesis

One potential computational mechanism that has been proposed to understand the cognitive processes of MBCT is the “Bayesian brain” hypothesis which suggests that the brain creates and continuously update a model of the external world (the physical environment and the physical self) based on input of past experience and homeostatic needs [29]. Neurophysiological studies have begun testing mechanisms proposed by the Bayesian brain hypothesis. For example, EEG studies have demonstrated that mathematical predictions of processes by Bayesian brain theories coincide with brain activity [30] [31] [32]. Predictive coding suggests how cognitive mechanisms are implemented physiologically, which allows for concrete and experimentally testable predictions for the neurophysiological processes affected by MBCT [33]. For example, changes in attentional focus change the precision weighting of sensory channels that are being attended to [34]. Thus, theories in predictive coding coincide with the training phase of MBCT, wherein the practitioner learns to maintain and re-orient attentional focus [35].

Processes within meditation training, such as MBCT training, can be understood from the perspective of Bayesian models of brain function by deriving experimentally testable predictions between attentional focus and sensory precision. Therefore, computational modeling becomes a clinical tool to obtain subject-specific estimates of precision-weighted prediction errors about physiological states.