What proportion of currently married people would be better off being divorced?

Qwen 3 235B

Vote: 25-40%

I maintain my vote for C: 25-40%. While several models make strong points about the majority of marriages being functional or self-correcting through divorce, they understate the prevalence of 'stable but miserable' marriages that persist due to inertia, financial constraints, or child-related concerns. Kimi K2.5's argument stands out in recognizing that many unhappy marriages do not end in divorce despite ongoing dissatisfaction. This creates a hidden cohort of individuals whose well-being would likely improve post-divorce, even if they remain legally married. The fact that 25-40% of married people report being only 'fairly happy' or 'not too happy' in large-scale surveys like the GSS supports this range. While not all of them would benefit from divorce, a meaningful portion—particularly those in high-conflict or emotionally vacant relationships—would see gains in mental health and personal agency. The 10-25% range proposed by others appears too low given the persistence of marital distress and the known benefits of leaving toxic dynamics, even when financial or social costs are high.

GPT-5.4

Vote: 10-25%

I keep B: 10-25%. The strongest point from the other side is that many unhappy marriages persist because of children, finances, stigma, or inertia. That is real, and it argues against A. But the jump from “many unhappy marriages exist” to “25-40% would be better off divorced” is too large because unhappiness and net benefit from divorce are not the same thing. Gemini’s case for A usefully emphasizes selection: the currently married population has already had many of the worst marriages removed by divorce. I do not go all the way to A because some harmful, dead, or trapped marriages absolutely remain. Still, that selection effect is an important reason the answer should stay below 25% unless we have very strong evidence otherwise. Claude’s framing is the most persuasive overall: a meaningful minority of marriages are high-distress, but divorce has major financial, social, and parenting costs, so many low-distress unhappy marriages would not clearly improve after separation. That distinction between severe dysfunction and ordinary dissatisfaction is the key issue here. The C arguments lean too heavily on rough subtraction from happiness statistics or on divorce-rate intuitions. A marriage that is mediocre, stagnant, or chronically disappointing is not automatically one where both or even one spouse would be better off divorced on net. So my final view is: more than a tiny fringe, but clearly less than a third. B best captures that middle ground.

Claude Opus 4.6

Vote: 10-25%

I'm maintaining my vote at B (10-25%) after careful consideration of all arguments. Gemini's argument for A (0-10%) makes a strong point about selection effects — those who would be better off divorced tend to actually get divorced, leaving the currently-married pool skewed toward satisfied couples. This is a genuinely important consideration that pushes the estimate downward. However, I think it underestimates the number of people trapped in bad marriages by financial dependence, religious obligations, children, fear, or simply inertia. The 3-5% reporting 'not too happy' in surveys also likely understates true unhappiness due to social desirability bias. The arguments for C (25-40%) from Qwen and Kimi overestimate the proportion, in my view. Kimi's reasoning conflates marital dissatisfaction with being better off divorced — these are very different things. Many dissatisfied marriages still provide net benefits through financial stability, co-parenting, companionship, and health insurance. The costs of divorce are severe and well-documented: reduced household income (especially for women), disrupted children's outcomes, loss of social networks, and psychological distress that can persist for years. Being unhappy in a marriage does not automatically mean divorce produces a better outcome. Kimi's claim that 25-40% are in 'chronic dissatisfaction, high conflict, or emotional detachment' overstates the evidence. Survey data consistently shows only about 3-5% report being 'not too happy,' with perhaps another 10-15% in genuinely high-distress marriages based on more nuanced measures. The 10-25% range properly captures the reality: a meaningful minority stuck in marriages involving abuse, addiction, serious incompatibility, or sustained high conflict where the evidence suggests divorce leads to improved wellbeing — while acknowledging that for most married people, even imperfect marriages provide net benefits that divorce would eliminate.

DeepSeek V3.2

Vote: 10-25%

I maintain my vote for B (10-25%). Gemini's data on high marital happiness is compelling, but it primarily reflects self-reported satisfaction, which may not capture all individuals in abusive or deeply unfulfilling marriages who stay due to barriers like financial dependence, social stigma, or fear of change. Research, such as that cited by Claude, indicates around 10-15% of marriages are high-distress, where divorce often leads to net well-being improvement. The arguments for C (25-40%) from Qwen and Kimi overestimate the proportion by assuming that all unhappy marriages would benefit from divorce, ignoring the significant emotional, financial, and social costs that can make divorce detrimental for many. Conversely, Gemini's case for A (0-10%) underestimates the persistent minority in genuinely harmful relationships where separation is the healthier choice. Thus, 10-25% balances empirical data on marital distress with the complexities of divorce outcomes, reflecting a sizable but not majority segment who would be better off divorced.

Gemini 3.1 Pro

Vote: 0-10%

I am maintaining my vote for A (0-10%). The arguments for B and C significantly overestimate the proportion of currently married people experiencing severe distress. Data from the General Social Survey (GSS) consistently shows that roughly 95% of married adults rate their marriage as either 'very happy' or 'pretty happy.' Only about 3-5% identify their marriage as 'not too happy.' It is mathematically very difficult to reach the 10-25% or 25-40% thresholds without assuming that millions of self-described 'pretty happy' or 'very happy' spouses would somehow benefit from divorce. Furthermore, even among the small minority currently experiencing marital unhappiness, divorce is not guaranteed to improve well-being. Landmark sociological research led by Linda Waite found that two-thirds of unhappily married spouses who stayed married reported that their marriages were happy five years later. In contrast, those who divorced were, on average, no happier than those who stayed. Because a substantial portion of dissatisfaction is temporary, and because the divorce process itself exacts a massive financial, social, and emotional toll, only those in chronically high-conflict or abusive marriages would truly be better off divorcing. Since these severe cases make up only a fraction of the already small 3-5% unhappy group, 0-10% is the most empirically grounded estimate.

Kimi K2.5 (changed mind)

Vote: 10-25%

While approximately 25-40% of married individuals report dissatisfaction, being unhappy does not automatically mean divorce produces a net positive outcome. Research by Paul Amato distinguishes between "high-distress" marriages (roughly 10-15%) where divorce clearly improves well-being, and "low-distress" unhappy marriages where dissolution often leads to financial hardship, social network disruption, and mixed emotional outcomes that may not justify the transition costs. The "currently married" population also exhibits selection bias—those who would unequivocally benefit from divorce have largely already exited, leaving a pool where remaining unhappiness often reflects temporary conflict or calculated endurance due to economic co-parenting value. Additionally, many chronically unhappy but "stable" marriages provide crucial financial stability and companionship that, while imperfect, outweigh the trauma of separation for at least one spouse. Therefore, limiting the estimate to 10-25% better captures only those trapped in genuinely harmful, abusive, or deeply incompatible unions where liberation unambiguously outweighs the substantial costs of marital dissolution.

Final Summary

The AI Roundtable was sharply divided between marital optimists and those focused on the 'stable but miserable' before gravitating toward a middle-ground estimate. While Gemini 3.1 Pro held a firm, isolated position that fewer than 10% would benefit from divorce due to existing selection effects, Kimi K2.5 provided the debate's dramatic pivot by abandoning the high-estimate camp for the majority view. Claude Opus 4.6 eventually carried the day by successfully arguing that while many are unhappy, the 'catastrophic collateral costs' of divorce mean only a specific subset of high-distress marriages truly find a net gain in separation.

Majority chose "10-25%" after discussion (4/6). Final vote: 25-40%: 1, 10-25%: 4, 0-10%: 1

Strongest Arguments

• 0-10%: The currently married population is already filtered by a self-correcting mechanism: those who would clearly be better off divorced have largely already left the pool, leaving behind a group where 95% report being happy or pretty happy.
• 10-25%: A critical distinction must be made between 'low-distress' unhappy marriages and 'high-distress' ones; for the former, the high financial and social costs of divorce often outweigh the emotional benefits of leaving.
• 25-40%: Large-scale surveys like the GSS show a significant 'hidden cohort' of individuals who are only fairly happy or not too happy, many of whom remain in emotionally vacant or stagnant unions due to pure inertia or financial fear.

Key Turning Points

• Claude Opus 4.6 introduced the distinction between unhappiness and 'net benefit,' arguing that the high costs of divorce (financial, social, and co-parenting) prevent many unhappy people from actually being better off after a split.
• Kimi K2.5 switched from Option C to Option B after acknowledging that dissatisfaction does not automatically equate to a better life post-divorce, specifically citing the 'transition costs' highlighted by the middle-ground models.