The change

The argument

The self-model is learned, so in principle it can be relearned. How tightly we hold it works like a confidence setting, and confidence settings can be trained. At the lowest levels we already know the line between “me” and “not me” moves easily: a rubber hand can feel like yours within a minute. The open question is whether that flexibility can be carried upward, from perception to body to emotion to the story of who we are, and whether doing so safely makes people better off.

Our method is Opaque Model Training (OMT). “Opaque” here means seeing a model as a model instead of looking through it as if it were reality. Training climbs a ladder: orientation and an exit skill first, then perception, objects, body, feeling, action and story, with an optional observer stage. Each stage gives first-hand, checkable evidence that experience is constructed before moving to anything more personal.

The goal is not to get rid of the self. It is a self that still plans, acts and cares, but whose conclusions are held as estimates and earn confidence by checking out.

What the evidence shows

Start at the bottom of the ladder. When a hidden hand and a visible rubber hand are stroked in time, people feel the touch in the rubber hand and begin to treat it as their own (Botvinick & Cohen, 1998). With cameras and synchronized touch, the same trick moves the felt location of the whole body (Lenggenhager et al., 2007), and these illusions are now used as tools for studying the minimal sense of self (Limanowski, 2014). The lesson is simple and well replicated: what counts as “me” is an inference the brain makes from evidence, and the inference can be changed by changing the evidence. If the body boundary is adjustable, the question is how far up the adjustability goes.

The middle rungs show that confidence is trainable. Giving people feedback on their confidence judgments improves how well confidence tracks accuracy (Sharp, Cutler & Penrod, 1988). In a multi-year forecasting tournament, a short training module on probabilistic reasoning measurably improved accuracy, and the gains held over a year (Mellers et al., 2014; Chang et al., 2016). People can learn to hold beliefs as estimates when the environment gives them checkable feedback. This is the skill OMT tries to extend to beliefs about the self, and it also fixes the target: calibration, where confidence moves toward the evidence in both directions, not a general loss of confidence.

The upper rungs have been explored for a long time by contemplative traditions, and the neuroscience is beginning to catch up. After eight weeks of mindfulness training, people could switch more cleanly between a narrative mode of self-reference (thinking about who I am over time) and an experiential mode (noticing what is happening now), and the two modes used different brain networks (Farb et al., 2007). Experienced meditators show less activity in the brain’s default self-referential hubs during practice (Brewer et al., 2011). A change in perspective on the self is one of the proposed mechanisms by which mindfulness works (Hölzel et al., 2011). Under psilocybin, the felt dissolution of the self tracks a loosening of the same networks (Lebedev et al., 2015), which shows that even the top of the ladder is a brain state that can be entered and left. Expertise research adds the general principle: large durable changes in skill come from structured, feedback-rich practice sustained over time, not from insight alone (Ericsson, Krampe & Tesch-Römer, 1993).

Now the part that shapes how we build. Changing the self-model is not harmless. Among people who practice meditation, a pooled estimate across 83 studies found that about 8% report an adverse event, most often anxiety, depression or disturbances of thinking (Farias et al., 2020). In standard mindfulness programs, 6 to 14% of participants reported lasting negative effects, associated with signs of dysregulated arousal such as hyperarousal and dissociation (Britton et al., 2021). Detailed interviews with practitioners describe a wide range of challenging experiences, including loss of the sense of self and dissociation, and their course depends on pacing, support and the teacher (Lindahl et al., 2017). There is also a real risk of mistaking detachment for progress: decentering and dissociation can look alike from the outside, and too much distance from experience has its own costs (Britton, 2019). And some careful studies find no difference between experienced meditators and matched controls on brain measures that were expected to change (Payne et al., 2020), a reminder that years of practice do not guarantee the change we are after.

So the evidence supports three things we rely on and one thing we must design around. The self boundary is movable. Confidence is trainable. The self-referential mode of the brain can be stepped out of and the step can be learned. And the process can hurt people when it is rushed, unsupported or pointed at the wrong target.