Measuring algorithmic bias to analyze the reliability of AI tools that predict depression risk using smartphone sensed-behavioral data

Adler, D. A., Stamatis, C. A., et al. (2024).

Npj Mental Health Research, 3(1), 17.

https://doi.org/10.1038/s44184-024-00057-y

Abstract

AI tools intend to transform mental healthcare by providing remote estimates of depression risk using behavioral data collected by sensors embedded in smartphones. While these tools accurately predict elevated depression symptoms in small, homogenous populations, recent studies show that these tools are less accurate in larger, more diverse populations. In this work, we show that accuracy is reduced because sensed-behaviors are unreliable predictors of depression across individuals: sensed-behaviors that predict depression risk are inconsistent across demographic and socioeconomic subgroups. We first identified subgroups where a developed AI tool underperformed by measuring algorithmic bias, where subgroups with depression were incorrectly predicted to be at lower risk than healthier subgroups. We then found inconsistencies between sensed-behaviors predictive of depression across these subgroups. Our findings suggest that researchers developing AI tools predicting mental health from sensed-behaviors should think critically about the generalizability of these tools, and consider tailored solutions for targeted populations.

Here are some thoughts:

This article presents a critical examination of the reliability and fairness of AI tools designed to predict depression risk using smartphone-sensed behaviors. The core finding is that these tools often fail to generalize across diverse populations because the relationship between behavior and depression is not universal. For instance, increased phone usage or changes in mobility may signal depression in one demographic group but not in another. This means that an AI model trained on one population can systematically underestimate or overestimate risk in another, turning algorithmic bias into a fundamental issue of measurement validity and psychometric reliability.

Importanlty, this underscores the necessity of approaching digital phenotyping tools with caution. Before adopting such technologies in clinical or screening contexts, it is vital to demand robust validation across the specific populations they intend to serve. The study also highlights tangible clinical risks: biased tools could misallocate mental health resources by overestimating risk in some groups while underestimating it in others—such as males, who already underutilize services. Ultimately, this research calls for a shift from aiming for universally generalizable models to developing targeted, culturally and contextually tailored solutions. Psychologists have a key role to play in this process by ensuring that digital tools are grounded in psychological theory, evaluated for equity, and implemented in a way that promotes ethical and effective mental health care for all.

The Supreme Court upholds the conviction of woman who challenged expert testimony in a drug case

Lindsay Whitehurst

apnews.com

originally posted 20 June 24

The Supreme Court on Thursday upheld the conviction of a California woman who said she did not know about a stash of methamphetamine hidden inside her car.

In a ruling that crossed the court’s ideological lines, the 6-3 majority opinion dismissed arguments that an expert witness for the prosecution had gone too far in describing the woman’s mindset when he said that most larger scale drug couriers are aware of what they are transporting.

“An opinion about most couriers is not an opinion about all couriers,” said Justice Clarence Thomas, who wrote the decision. He was joined by fellow conservatives Chief Justice John Roberts, Justices Samuel Alito, Brett Kavanaugh and Amy Coney Barrett as well as liberal Justice Ketanji Brown Jackson.

In a sharp dissent, conservative Justice Neil Gorsuch wrote that the ruling gives the government a “powerful new tool in its pocket.”

“Prosecutors can now put an expert on the stand — someone who apparently has the convenient ability to read minds — and let him hold forth on what ‘most’ people like the defendant think when they commit a legally proscribed act. Then, the government need do no more than urge the jury to find that the defendant is like ‘most’ people and convict,” he wrote. Joining him were the court’s other liberal justices, Sonia Sotomayor and Elena Kagan.

The info is here.

Here are some thoughts:

The recent Supreme Court case involving a woman convicted of drug trafficking highlights a complex issue surrounding expert testimony, particularly for psychologists. In this case, the prosecution's expert offered an opinion on the general awareness of large-scale drug couriers, which the defense argued unfairly portrayed the defendant's mindset. While the Court allowed the testimony, it leaves some psychologists concerned.

The potential for expert testimony to blur the lines between general patterns and specific defendant behavior is a worry. Psychologists strive to present nuanced assessments based on individual cases. This ruling might incentivize broader generalizations, which could risk prejudicing juries against defendants. It's crucial to find a balance between allowing experts to provide helpful insights and ensuring they don't overstep into determining a defendant's guilt.

Moving forward, psychologists offering expert testimony may need to tread carefully.  They should ensure their testimony focuses on established psychological principles and avoids commenting on a specific defendant's knowledge or intent. This case underscores the importance of clear guidelines for expert witnesses to uphold the integrity of the justice system.

Divide and Rule? Why Ethical Proliferation is not so Wrong for Technology Ethics.

Llorca Albareda, J., Rueda, J.

Philos. Technol. 36, 10 (2023).

https://doi.org/10.1007/s13347-023-00609-8

Abstract

Although the map of technology ethics is expanding, the growing subdomains within it may raise misgivings. In a recent and very interesting article, Sætra and Danaher have argued that the current dynamic of sub-specialization is harmful to the ethics of technology. In this commentary, we offer three reasons to diminish their concern about ethical proliferation. We argue first that the problem of demarcation is weakened if we attend to other sub-disciplines of technology ethics not mentioned by these authors. We claim secondly that the logic of sub-specializations is less problematic if one does adopt mixed models (combining internalist and externalist approaches) in applied ethics. We finally reject that clarity and distinction are necessary conditions for defining sub-fields within ethics of technology, defending the porosity and constructive nature of ethical disciplines.

Conclusion

Sætra and Danaher have initiated a necessary discussion about the increasing proliferation of neighboring sub-disciplines in technology ethics. Although we do not share their concern, we believe that this debate should continue in the future. Just as some subfields have recently been consolidated, others may do the same in the coming decades. The possible emergence of novel domain-specific technology ethics (say Virtual Reality Ethics) suggests that future proposals will point to as yet unknown positive and negative aspects of this ethical proliferation. In part, the creation of new sub-disciplines will depend on the increasing social prominence of other emerging and future technologies. The map of technology ethics thus includes uncharted waters and new subdomains to discover. This makes ethics of technology a fascinatingly lively and constantly evolving field of knowledge.

The info is here.

The globalizability of temporal discounting

Ruggeri, K., Panin, A., et al. (2021, October 1). 

psyarxiv.com

https://doi.org/10.31234/osf.io/2enfz

Abstract

Economic inequality is associated with extreme rates of temporal discounting, which is a behavioral pattern where individuals choose smaller, immediate financial gains over larger, delayed gains. Such patterns may feed into rising global inequality, yet it is unclear if they are a function of choice preferences or norms, or rather absence of sufficient resources to meet immediate needs. It is also not clear if these reflect true differences in choice patterns between income groups. We test temporal discounting and five intertemporal choice anomalies using local currencies and value standards in 61 countries. Across a diverse sample of 13,629 participants, we found highly consistent rates of choice anomalies. Individuals with lower incomes were not significantly different, but economic inequality and broader financial circumstances impact population choice patterns.

Technical Abstract

Economic inequality is associated with extreme rates of temporal discounting, which is a behavioral pattern  where  individuals  choose  smaller,  immediate  financial  gains  over larger, delayed gains. Such patterns may feed into rising global inequality, yet it is unclear if  they are a function of choice preferences or norms, or rather absence of sufficient resources to meet immediate needs. It is also not clear if these reflect true differences in choice  patterns  between  income  groups.  We  test  temporal  discounting and  five intertemporal choice anomalies using local currencies and value standards in 61 countries. Across a diverse sample of 13,629 participants, we found highly consistent rates choice anomalies. Individuals with lower incomes were not significantly different, but economic inequality and broader financial circumstances impact population choice patterns.

The research is here.

Bottom line: This research refutes the perspective that low-income individuals are poor decision-makers.