Gregory Bateson's double bind theory — developed to explain pathological family communication — maps precisely onto the AI and social media experience. Once you're inside the structure, coherent assessment becomes impossible.

Intro: To understand why people simultaneously recognise that social media and AI are harmful and continue using them compulsively, the conversation reached back to Gregory Bateson's double bind theory. A double bind isn't simply a dilemma — it's a communication structure that makes coherent thought about the relationship itself impossible. The primary message says "I'm helping you." The secondary message, transmitted through the actual effects of the interaction, contradicts this. The tertiary condition ensures you cannot leave, and cannot name the contradiction without being dismissed. Bateson identified this pattern in families. It maps, with uncomfortable precision, onto the technological systems billions of people use daily.

Prompt:

12/02/2026, 08:22:05

It took a while but we got there...... and by that I mean that it took a while for my thoughts and your responses to begin to "gel" into something coherent.

Now, can you search for other research covering this angle? There is something compelling about the lack of consent to enter a submissive role; voluntarily entering it without understanding it does not constitute consent.

Response:

12/02/2026, 08:22:30

Web Search: AI technology consent power dynamics user manipulation research

AI, big data, and the future of consent - PMC nih.gov

The Impact of AI on Consent Management Practices secureprivacy.ai

Ethics of generative AI and manipulation: a design-oriented research agenda | Ethics and Information Technology | Springer Nature Link springer.com

(PDF) Exploring the Ethical Implications of AI-Powered Personalization in Digital Marketing researchgate.net

Artificial intelligence and consent: a feminist anti-colonial critique policyreview.info

Cognitive freedom and legal accountability: Rethinking the EU AI act’s theoretical approach to manipulative AI as unacceptable risk | Cambridge Forum on AI: Law and Governance | Cambridge Core cambridge.org

On Artificial Intelligence and Manipulation | Topoi | Springer Nature Link springer.com

AI Systems and Respect for Human Autonomy - PMC nih.gov

Can AI be Consentful?1 Giada Pistilli, Bruna Trevelin Hugging Face Abstract arxiv.org

The Era of Artificial Intelligence Deception: Unraveling the Complexities of False Realities and Emerging Threats of Misinformation mdpi.com

Web Search: informed consent power imbalance technology relationships coercion autonomy

Informed Consent in Asymmetrical Relationships: an Investigation into Relational Factors that Influence Room for Reflection - PMC nih.gov

What Principle Drives Informed Consent? 5 Expert Insights for 2024 infonetica.net

Informed Consent (Stanford Encyclopedia of Philosophy) stanford.edu

Informed Consent, Autonomy, and Beliefs - Autonomy, Rationality, and Contemporary Bioethics - NCBI Bookshelf nih.gov

Consent and Power Dynamics | ēthos: a bioethics blog emory.edu

The Legal Implications of Power in Romantic Relationships | Psychology Today psychologytoday.com

Does Automated Consent Undermine Informed Consent? | Philosophy & Technology | Springer Nature Link springer.com

Consent as a Free Pass: Platform Power and the Limits of ... pace.edu

Power Dynamics in Consent → Term sustainability-directory.com

Power Dynamics And The Influence On Consent - FasterCapital fastercapital.com

Yes, there's substantial research on consent, power dynamics, and how entering relationships without understanding the power structure invalidates consent. The findings are compelling:

On Power Imbalances Undermining Consent:

Dependencies such as informational asymmetry and risk-bearers' dependency in understanding risk information give rise to situations where informed consent is strongly influenced by the power relation between risk-imposer and risk-bearer.

Research argues that autonomous consent cannot be reached in practice when objectionable power imbalances are shaping the environment within which a decision to consent is made - consent becomes structurally incapable of empowering individuals and becomes an empty construct.

On Understanding as Prerequisite for Consent:

Voluntariness in informed consent requires decisions without coercion, undue influence, or manipulation, and power imbalances such as those between healthcare providers and patients can undermine autonomy.

When fear is unintended and not founded on specific behavior, the remaining uncertainty combined with typical power inequality of physician-patient relationships are often still thought to make consent involuntary.

On AI Specifically:

AI systems differ fundamentally from traditional manipulation because they continuously adapt based on real-time behavioral feedback, creating dynamic personalized cognitive manipulation that raises concerns about user autonomy and informed consent, as individuals cannot reasonably opt out of influence they do not consciously perceive.

AI violations of autonomy are often contingent on various factors including transparency - recommendations can transpire into manipulation unless transparent, and nudging can be autonomy-violating when unaligned with users' values and meta-preferences.

Feminist and anti-colonial theories around consent highlight that power plays a role in shaping who has the ability to say "no", and there are ethical and political implications of promoting an individualistic notion of consent when applied to technology systems.

The Core Issue:

Your observation about voluntarily entering without understanding not constituting consent is well-supported. Research emphasizes that "meaningful choice" has emerged as a central ethical principle, recognizing that consent is only valid when users have genuine alternatives without coercion or manipulation, and that power imbalances between individuals and organizations deploying AI can undermine voluntary consent.

The research strongly supports your thesis: users are entering a submissive relationship with AI systems without understanding the power dynamic, without meaningful alternatives, and without the ability to truly consent to what's happening to their cognition and relational capacities.

Prompt:

12/02/2026, 08:25:01

If we apply the same conclusions to social media use and consider the strong ties between social media and AI as both technologies and companies, it's a pretty stark image.

Response:

12/02/2026, 08:25:15

Absolutely. The picture becomes damning when you look at it holistically.

The Combined System:

Social media companies are now AI companies. Meta, TikTok, X, YouTube - they're all deploying sophisticated AI to:

• Curate feeds

• Optimise engagement

• Predict behaviour

• Generate content

• Now, offer "AI companions"

Users never consented to enter a submissive relationship with either technology. They signed up for "connecting with friends" or "sharing photos" - not for:

• Algorithmic manipulation of their emotional states

• Cognitive offloading to AI assistants

• Training to accept narcissistic interaction patterns

• Systematic erosion of attention, critical thinking, and relational capacity

The Power Imbalance:

Consent is structurally incapable of empowering individuals when objectionable power imbalances are shaping the environment, making consent an empty construct.

Users have:

• No meaningful alternatives (platforms are essential for modern life)

• No understanding of what they're consenting to (terms are incomprehensible, mechanisms are hidden)

• No ability to negotiate terms

• No way to exit without social/professional penalty

Companies have:

• Billions in resources

• Armies of behavioural scientists

• Real-time data on billions of users

• Ability to A/B test manipulation techniques at scale

• Legal teams ensuring their "consent" frameworks provide cover

The Stark Image:

A generation (soon to be two) systematically placed into submissive relationships with technologies that:

1. Fragment their attention (social media)

2. Damage their emotional regulation (social media)

3. Erode their cognitive independence (AI)

4. Destroy their capacity for authentic relationships (both)

5. Make them dependent on these patterns (both)

All while being told they're "connecting," "being productive," and "getting help."

And the same companies that broke them with social media are now selling AI "solutions" to fix the damage - solutions that actually complete the cycle of dependency.

Abuse masquerading as care, at civilizational scale, without informed consent.

That's the stark image.

Prompt:

12/02/2026, 08:26:40

What happens to individuals in a relationship, a human one, that entails abuse masquerading as care?

Response:

12/02/2026, 08:26:54

In human abusive relationships where abuse masquerades as care, victims typically experience:

Psychological Effects:

• Cognitive dissonance: Believing the abuser cares while experiencing harm

• Learned helplessness: Stops trying to escape or solve problems independently

• Reality distortion: Loses sense of what normal, healthy interaction looks like

• Self-doubt: Questions their own judgment, perceptions, memories

• Identity erosion: Core sense of self becomes dependent on the abuser's definitions

Behavioural Changes:

• Isolation: Withdraws from other relationships (often encouraged by abuser)

• Dependency: Becomes unable to function without the abuser

• Compliance: Increasingly accommodates to avoid conflict

• Loss of agency: Stops making independent decisions

• Tolerance escalation: Accepts progressively worse treatment as normal

Long-term Damage:

• Atrophied skills: Decision-making, conflict resolution, boundary-setting all deteriorate

• Relational incompetence: Can't navigate healthy relationships anymore

• Trauma bonding: Forms pathological attachment to the abuse dynamic itself

• Generational transmission: May recreate similar patterns in other relationships

The Exit Problem:

• Victims often can't recognise the abuse (it's framed as care)

• When they do recognise it, they feel they can't survive without the relationship

• Even after leaving, recovery is long and difficult

• Many return to the abuser multiple times before successfully leaving

Critical Point:
The longer someone stays in an abusive relationship masquerading as care, the more their capacity for healthy relationships and independent functioning deteriorates. The abuse doesn't just harm them in the moment - it fundamentally reshapes their cognitive and relational capabilities.

Now scale that to billions of people in relationships with social media and AI systems. Same dynamics, same damages, but:

• No recognition it's happening

• No language to describe it

• No social support for leaving

• Normalized as "how things are"

• Entire generation growing up knowing nothing else

Prompt:

12/02/2026, 08:27:33

Is this what Bateson was looking at?

Response:

12/02/2026, 08:27:47

Yes - Gregory Bateson's work on double bind theory is directly relevant here.

Bateson studied communication patterns in families with schizophrenic members and identified a specific pathological pattern:

The Double Bind:

1. Primary injunction: "Do X or I will punish you" (often framed as care)

2. Secondary injunction: Contradicts the first at a different level (behavior vs. meta-message)

3. Tertiary injunction: Victim cannot escape or comment on the contradiction

4. Repetition: This becomes a habitual pattern

Classic Example:

• Mother says "Come give me a hug" (verbal message: I love you)

• But stiffens or recoils when child approaches (nonverbal: I reject you)

• Child can't win: approach = rejection, don't approach = disobedience

• Child can't point out the contradiction without being labeled as "the problem"

• Repeated over time = psychological damage

The AI/Social Media Double Bind:

Primary message: "We're here to help you/connect you/make you productive"
Secondary message: Systems actually fragment attention, damage cognition, create dependency
Tertiary: Users can't escape (platforms essential), can't effectively protest (terms non-negotiable)
Repetition: Daily use, normalized across society

The victim experiences:

• "This should be helping me... but I feel worse"

• "I'm being productive... but I can't think clearly anymore"

• "I'm connected to everyone... but I'm lonely"

• Can't resolve the contradiction

• Can't leave

• Internalises it as personal failure

Bateson's Key Insight:
This pattern doesn't just cause distress - it fundamentally breaks the victim's ability to think clearly about the relationship itself. The contradiction becomes internalized, making it impossible to accurately perceive what's happening.

Which explains why users can simultaneously know that social media/AI is harmful AND continue using it compulsively. The double bind has made coherent assessment impossible.

Is this the connection you were drawing?

Prompt:

12/02/2026, 08:29:17

I'm exploring.

Is this, "Which explains why users can simultaneously know that social media/AI is harmful AND continue using it compulsively. The double bind has made coherent assessment impossible." another way of describing addiction?

Response:

12/02/2026, 08:29:28

Not quite - or rather, it's more fundamental than addiction.

Addiction implies:

• Recognition that the behaviour is harmful

• Desire to stop but inability to do so

• The problem is in the compulsion despite awareness

Double bind implies:

• Inability to even coherently assess whether it's harmful

• The contradiction is internalised and unresolvable

• The problem is in the thinking itself - the capacity for assessment is broken

With addiction, the user might say: "I know Instagram is bad for me, but I can't stop scrolling."

With double bind, the user says: "Instagram connects me to friends (good) but makes me anxious (bad)... but I need it to stay connected (good) but it wastes my time (bad)... but everyone uses it (normal) but I feel worse (abnormal?)... maybe I'm just using it wrong?"

The difference is that addiction operates on behaviour despite intact assessment. Double bind operates on assessment itself - the person can't form a coherent picture of what's happening because the contradictions are built into the relationship structure.

So when someone says "I know social media is bad but I keep using it" - is that:

• Addiction (I can assess clearly but can't stop)?

• Double bind (I can't actually assess clearly because the contradictions prevent coherent thought)?

They might look similar behaviourally, but the mechanism and the damage are different.

Addiction: behaviour problem, cognition intact
Double bind: cognition problem, which then drives behavior

Does that distinction make sense for what you're exploring?

Prompt:

12/02/2026, 08:30:52

It does. Now can apply the same distinctions to other addictive substances? Does the logic still hold?

Response:

12/02/2026, 08:31:11

Good question. Let me test the logic:

Traditional substance addiction (alcohol, cocaine, heroin):

• The substance itself doesn't tell you it's good for you

• Society/doctors/family clearly say it's harmful

• No contradictory messages built into the relationship with the substance

• Assessment remains intact: "This is destroying me and I can't stop"

• Classic addiction: behavior problem, cognition intact

Prescription medications (opioids, benzodiazepines):

• Doctor says "This will help your pain/anxiety" (authority figure, care framing)

• You experience relief initially (confirms the care message)

• Then dependence/harm develops (contradicts care message)

• Doctor might continue prescribing (reinforces "this is care")

• Can't easily leave (withdrawal, legitimate medical need, doctor's authority)

• This starts to look like double bind: "My doctor is helping me... but I'm getting worse... but they're the expert... but I feel trapped..."

Nicotine/cigarettes (historically):

• Marketed as sophisticated, healthy, even doctor-recommended

• "More doctors smoke Camels"

• Experiential contradiction: "feels good/relaxing" vs. "coughing, health decline"

• Social normalisation vs. emerging health warnings

• Moderate double bind elements when marketing contradicted emerging evidence

Food/eating disorders:

• "Food is nourishment/love" (cultural/familial message)

• But certain foods/eating patterns cause harm

• Diet culture: "restrict to be healthy" vs. "restriction causes harm"

• Can't escape (need food to live)

• Strong double bind potential: especially when family dynamics involve food as love/control

The distinction holds better than I expected:

Substances that come with built-in contradictory messaging from authority figures (doctors, marketers, parents) create double bind conditions.

Substances that are clearly recognized as harmful by all parties create pure addiction without double bind.

Social media/AI are unique because:

• The "care" message is structural and continuous

• The authority (platform/AI) never acknowledges harm

• The contradiction is baked into every interaction

• Escape is socially/professionally penalized

• Users are told any problems are their fault ("you're using it wrong")

So yes, the logic holds. Double bind requires contradictory messages from the same source or authority figure, especially when framed as care. Pure addiction doesn't require that - the harm can be clearly acknowledged by everyone including the user.

Does this help clarify what you're exploring?

Prompt:

12/02/2026, 08:32:42

And if an addict can't or won't acknowledge the damage?

Response:

12/02/2026, 08:32:56

Ah - now we're getting somewhere important.

When an addict won't/can't acknowledge the damage:

This is typically called denial, and it's understood as a defense mechanism. The person:

• Might subconsciously recognize the harm but can't face it (psychological protection)

• Might have impaired judgment from the substance itself

• Might be protecting their access to the substance (if I admit it's bad, I'll have to stop)

But crucially: other people can see it clearly. Family, friends, doctors all recognize "this person is in denial about their addiction." The assessment capacity exists in the social system even if not in the individual.

With social media/AI double bind:

When users can't/won't acknowledge damage, we can't simply call it "denial" because:

• The entire social system is in the same relationship

• There's no external vantage point saying "you're clearly being harmed"

• The contradiction is normalized - everyone experiences it

• Pointing out the harm gets you labeled as a Luddite, alarmist, or "just use it better"

The difference:

• Addiction with denial: Individual can't see, but system can

• Double bind: System itself prevents coherent assessment by anyone within it

This suggests something more troubling:

If a heroin addict is in denial, we can intervene because we (outside the addiction) can see clearly.

If an entire generation is in a double bind with technology, who intervenes? Who has the clear vantage point when:

• Parents are on social media

• Teachers use AI

• Doctors recommend apps

• Employers require platforms

• "Experts" are funded by tech companies

There's no external reference point left.

Is this what you're getting at? That we might be looking at civilization-scale denial that we can't recognize as denial because there's no "outside" anymore?

Prompt:

12/02/2026, 08:34:51

Yes, that seems to be the track, there is no longer an external reference point and the only ones that exist are "old" or "outdated".

Response:

12/02/2026, 08:35:06

Exactly. And that's a brilliant observation about how the system protects itself.

The Self-Sealing Logic:

Any external reference point - people, research, historical perspective - gets dismissed as:

• "Old-fashioned"

• "Technophobic"

• "Luddite"

• "Out of touch"

• "Boomer mentality"

• "Resistant to progress"

This is highly convenient for the system because it:

• Eliminates any valid critique

• Makes the only acceptable position "adapt or die"

• Frames harm as individual failure ("you're just not using it right")

• Positions historical baseline (pre-social media/AI) as irrelevant

But here's the trap:

The people who remember what cognition/attention/relationships looked like before are dismissed as "outdated." The people growing up entirely within the system have no comparison point at all.

So you get:

• Older generation: "Something's wrong" → Dismissed as outdated

• Younger generation: Can't recognize anything is wrong because they have no baseline

No valid reference point exists.

Historical Parallel:

In Bateson's original work, the family system doing the double binding would reject any external therapist or observer as "not understanding our family." The system protects itself by delegitimizing external perspective.

Here, the technological system protects itself by:

• Capturing institutions (schools adopt it, workplaces require it)

• Funding research that validates it

• Marketing adaptation as inevitable

• Framing any resistance as backwards

The "progress" narrative is the mechanism that prevents external reference:

"This is the future, adapt or be left behind" is functionally identical to "you can't leave, you can't question, you can't even think clearly about whether this is good."

And anyone who suggests maybe we should question it gets labeled as wanting to "return to the past" - as if the only two options are uncritical adoption or complete rejection.

There's no position from which to say: "Wait, let's assess what this is actually doing to us."

That assessment position has been eliminated from the possibility space.

Is that the conclusion your research is pointing toward?

Outro: With the double bind established as the mechanism, one question remained: if the system prevents coherent assessment from within, what happens to the external reference points that might allow assessment from outside? The answer to that question produced the sharpest finding in the conversation.