Sathyaraj Asaithambi

Enterprise skills graphs, which connect skills to the people who hold them and the work that requires them, have become the data substrate on which agentic HR systems recommend hiring, mobility, reskilling, and redeployment. A consensus has formed among HR analysts that the value of these systems depends less on the sophistication of the agents than on the trustworthiness of the graph beneath them. That trustworthiness, however, is discussed almost entirely in qualitative terms. Where it is quantified at all, it is reduced to a single proxy: coverage, the percentage of roles or people for which some skill data exists. Coverage answers how much data an organisation holds. It is silent on whether that data can be trusted enough to let an autonomous agent act. This paper develops the missing measurement science. Working within a design-science method and drawing on established data-quality standards (ISO/IEC 25012, ISO 8000, the FAIR principles) and on machine-learning calibration theory, it defines a reliability construct for skills graphs along five dimensions: Freshness, Accuracy of stated confidence, Contestability, Traceability, and Stability, which form the mnemonic FACTS. The dimensions are derived from the conditions under which an automated decision is defensible rather than assembled as a list, and each is operationalised as a metric computable on a graph's nodes and edges. They are combined into a dashboard-ready Skills Graph Reliability Index (SGRI) through a gated, deny-overrides aggregation rather than a weighted mean, so that weakness on any one dimension cannot be concealed by strength on another. SGRI resolves to three decision bands: act, human-gate, or do not automate. The paper makes five contributions. To begin with, it argues that unreliability in a skills graph compound rather than averages out under agentic use, because agent inferences written back without provenance contaminate the evidence against which later inferences would be validated, a mechanism named here evidential circularity. Second, it demonstrates that mechanism computationally. In a simulation of 20,000 edges over twelve cycles, unlabelled write-back caused measured calibration error to improve while true calibration error degraded, and the divergence increased with the write-back rate. At a write-back rate of 0.50 the contaminated graph reported a lower calibration error than a fully labelled control while its true error was approximately 4.8 times worse, so an organisation comparing the two on measured calibration would have preferred the corrupted graph. Third, the paper specifies the SGRI aggregation and decision bands and defends the deny-overrides rule on grounds of asymmetric loss. Fourth, it states the boundary of the construct explicitly: SGRI measures the trustworthiness of the evidence base, not the quality of inference drawn from it. Fifth, it confronts the ground-truth problem, since calibration in this domain depends on an oracle that is itself biased, and proposes triangulation and mandatory oracle-provenance disclosure as partial remedies. Fairness is treated as a stratification requirement across all five dimensions rather than as a sixth dimension. A reference architecture, a worked illustration, and a set of falsifiable propositions complete the paper. No claim is made that the index predicts decision quality; that relationship is stated as a falsifiable proposition and remains untested.

Enterprise skills graphs, which connect skills to the people who hold them and the work that requires them, have become the data substrate on which agentic HR systems recommend hiring, mobility, reskilling, and redeployment. A consensus has formed among HR analysts that the value of these systems depends less on the sophistication of the agents than on the trustworthiness of the graph beneath them. That trustworthiness, however, is discussed almost entirely in qualitative terms. Where it is quantified at all, it is reduced to a single proxy: coverage, the percentage of roles or people for which some skill data exists. Coverage answers how much data an organisation holds. It is silent on whether that data can be trusted enough to let an autonomous agent act. This paper develops the missing measurement science. Working within a design-science method and drawing on established data-quality standards (ISO/IEC 25012, ISO 8000, the FAIR principles) and on machine-learning calibration theory, it defines a reliability construct for skills graphs along five dimensions: Freshness, Accuracy of stated confidence, Contestability, Traceability, and Stability, which form the mnemonic FACTS. The dimensions are derived from the conditions under which an automated decision is defensible rather than assembled as a list, and each is operationalised as a metric computable on a graph's nodes and edges. They are combined into a dashboard-ready Skills Graph Reliability Index (SGRI) through a gated, deny-overrides aggregation rather than a weighted mean, so that weakness on any one dimension cannot be concealed by strength on another. SGRI resolves to three decision bands: act, human-gate, or do not automate. The paper makes five contributions. To begin with, it argues that unreliability in a skills graph compound rather than averages out under agentic use, because agent inferences written back without provenance contaminate the evidence against which later inferences would be validated, a mechanism named here evidential circularity. Second, it demonstrates that mechanism computationally. In a simulation of 20,000 edges over twelve cycles, unlabelled write-back caused measured calibration error to improve while true calibration error degraded, and the divergence increased with the writeback rate. At a write-back rate of 0.50 the contaminated graph reported a lower calibration error than a fully labelled control while its true error was approximately 4.8 times worse, so an organisation comparing the two on measured calibration would have preferred the corrupted graph. Third, the paper specifies the SGRI aggregation and decision bands and defends the deny-overrides rule on grounds of asymmetric loss. Fourth, it states the boundary of the construct explicitly: SGRI measures the trustworthiness of the evidence base, not the quality of inference drawn from it. Fifth, it confronts the ground-truth problem, since calibration in this domain depends on an oracle that is itself biased, and proposes triangulation and mandatory oracle-provenance disclosure as partial remedies. Fairness is treated as a stratification requirement across all five dimensions rather than as a sixth dimension. A reference architecture, a worked illustration, and a set of falsifiable propositions complete the paper. No claim is made that the index predicts decision quality; that relationship is stated as a falsifiable proposition and remains untested.

The EU Artificial Intelligence Act (Regulation (EU) 2024/1689) designates AI systems used in employment, worker management, and access to self-employment as high-risk under Annex III. Such systems attract stringent obligations for risk management, data governance, transparency, human oversight, recordkeeping, and conformity assessment. A fast-growing class of enterprise systems sits ambiguously against this category: external workforce-intelligence platforms that convert public labour-market signals, such as job postings, competitor movements, and skill-demand shifts, into aggregate strategic guidance. These platforms inform workforce planning at the level of roles, functions, and populations rather than making decisions about identifiable individuals. The boundary between high-risk employment AI and governed strategic decision support is, however, neither self-evident nor consistently applied in practice. This paper develops a classification framework that distinguishes workforce-intelligence architectures falling within the Act's high-risk employment scope from those that constitute defensible decision support. Following the design-science research paradigm, and drawing on the derogation criteria in Article 6(3), it analyses four system archetypes and identifies five architectural properties that materially affect classification and defensibility: an aggregation boundary, deterministic and reproducible processing, explain-only outputs, human-in-the-loop escalation, and end-to-end provenance logging. The properties are instantiated on a generalised reference architecture, translated into an operational compliance checklist, and extended into a practice layer that covers vendor and product assurance, effectiveness and bias testing across the lifecycle, post-deployment monitoring and re-classification triggers, and an operating model of roles with worker-representative consultation. The paper offers HR leaders, system designers, and governance functions a practical instrument for the Act's phased enforcement, and sets out a research agenda for responsible workforce intelligence.

The EU Artificial Intelligence Act (Regulation (EU) 2024/1689) designates AI systems used in employment, worker management, and access to self-employment as high-risk under Annex III. Such systems attract stringent obligations for risk management, data governance, transparency, human oversight, record-keeping, and conformity assessment. A fast-growing class of enterprise systems sits ambiguously against this category: external workforce-intelligence platforms that convert public labour-market signals, such as job postings, competitor movements, and skill-demand shifts, into aggregate strategic guidance. These platforms inform workforce planning at the level of roles, functions, and populations rather than making decisions about identifiable individuals. The boundary between high-risk employment AI and governed strategic decision support is, however, neither self-evident nor consistently applied in practice. This paper develops a classification framework that distinguishes workforce-intelligence architectures falling within the Act's high-risk employment scope from those that constitute defensible decision support. Following the design-science research paradigm, and drawing on the derogation criteria in Article 6(3), it analyses four system archetypes and identifies five architectural properties that materially affect classification and defensibility: an aggregation boundary, deterministic and reproducible processing, explain-only outputs, human-in-the-loop escalation, and end-to-end provenance logging. The properties are instantiated on a generalised reference architecture, translated into an operational compliance checklist, and extended into a practice layer that covers vendor and product assurance, effectiveness and bias testing across the lifecycle, post-deployment monitoring and re-classification triggers, and an operating model of roles with worker-representative consultation. The paper offers HR leaders, system designers, and governance functions a practical instrument for the Act's phased enforcement, and sets out a research agenda for responsible workforce intelligence.

Background: Ocular pain has multifactorial etiologies that affect activities of daily life, psychological well-being, and health-related quality of life (QoL). Chronic ocular surface pain (COSP) is a persistent eye pain symptom lasting for a period longer than 3 months. Objective: The objective of this social media listening study was to better understand COSP and related symptoms and identify its perceived causes, comorbidities, and impact on QoL from social media posts. Methods: A search from February 2020 to February 2021 was performed on social media platforms (Twitter, Facebook, blogs, and forums) for English-language content posted on the web. Social media platforms that did not provide public access to information or posts were excluded. Social media posts from Australia, Canada, the United Kingdom, and the United States were retrieved using the Social Studio platform—a web-based aggregator tool. Results: Of the 25,590 posts identified initially, 464 posts about COSP were considered relevant; the majority of conversations (98.3%, n=456) were posted by adults (aged >18 years). Work status was mentioned in 52 conversations. Patients’ or caregivers’ discussions across social media platforms were centered around the symptoms (61.9%, n=287) and causes (58%, n=269) of ocular pain. Patients mentioned having symptoms associated with COSP, including headache or head pressure, dry or gritty eyes, light sensitivity, etc. Patients posted that their COSP impacts day-to-day activities such as reading, driving, sleeping, and their social, mental, and functional well-being. Conclusions: Insights from this study reported patients’ experiences, concerns, and the adverse impact on overall QoL. COSP imposes a significant burden on patients, which spans multiple aspects of daily life.

Background: Graft-versus-host disease (GVHD) is the major cause of short- and long-term morbidity and mortality after allogeneic hematopoietic stem cell transplantation. Treatment options beyond corticosteroid therapy remain limited, and prolonged treatment often leads to impaired quality of life (QoL). A better understanding of the needs and experiences of patients with GVHD is required to improve patient care. Objective: The aim of this study is to explore different social media (SM) channels for gathering and analyzing the needs and experiences of patients and other stakeholders across 14 European countries. Methods: We conducted a retrospective analysis of SM data from the public domain. The Talkwalker social analytics tool collected data from open-access forums, blogs, and various social networking sites using predefined search strings. The raw data set derived from the aggregator tool was automatically screened for the relevancy of posts, generating the curated data set that was manually reviewed to identify posts that fell within the predefined inclusion and exclusion criteria. This final data set was then used for the deep-dive analysis. Results: A total of 9016 posts relating to GVHD were identified between April 2019 and April 2021. Deduplication and relevancy checks resulted in 325 insightful posts, with Twitter contributing 250 (77%) posts; blogs, 49 (15%) posts; forums, 13 (4%) posts; Facebook, 7 (2%) posts; and Instagram and YouTube, 4 (1%) posts. Patients with GVHD were the primary stakeholders, contributing 63% of all SM posts. In 234 posts, treatment was the most discussed stage of the patient journey (68%), followed by symptoms (33%), and diagnosis and tests (21%). Among treatment-related posts (n=159), steroid therapy was most frequently reported (54/159, 34%). Posts relating to treatment features (n=110) identified efficacy (45/110, 41%), side effects (38/110, 35%), and frequency and dosage (32/110, 29%), as the most frequently discussed features. Symptoms associated with GVHD were described in 24% (77/325) of posts, including skin-related conditions (49/77, 64%), dry eyes or vision change (13/77, 17%), pain and cramps (16/77, 21%), and fatigue or muscle weakness (12/77, 16%). The impacts of GVHD on QoL were discussed in 51% (165/325) of all posts, with the emotional, physical and functional, social, and financial impacts mentioned in 69% (114/165), 50% (82/165), 5% (8/165), and 2% (3/165) of these posts, respectively. Unmet needs were reported by patients or caregivers in 24% (77/325) of analyzed conversations, with treatment-related side effects being the most common (35/77, 45%) among these posts. Conclusions: SM listening is a useful tool to identify medical needs. Treatment of GVHD, including treatment-related side effects, as well as its emotional and physical impact on QoL, are the major topics that GVHD stakeholders mention on SM. We encourage a structured discussion of these topics in interactions between health care providers and patients with GVHD. Trial Registration: Not applicable

BACKGROUND Ocular pain has multifactorial etiologies that affect activities of daily life, psychological well-being, and health-related quality of life (QoL). Chronic ocular surface pain (COSP) is a persistent eye pain symptom lasting for a period longer than 3 months. OBJECTIVE The objective of this social media listening study was to better understand COSP and related symptoms and identify its perceived causes, comorbidities, and impact on QoL from social media posts. METHODS A search from February 2020 to February 2021 was performed on social media platforms (Twitter, Facebook, blogs, and forums) for English-language content posted on the web. Social media platforms that did not provide public access to information or posts were excluded. Social media posts from Australia, Canada, the United Kingdom, and the United States were retrieved using the Social Studio platform—a web-based aggregator tool. RESULTS Of the 25,590 posts identified initially, 464 posts about COSP were considered relevant; the majority of conversations (98.3%, n=456) were posted by adults (aged >18 years). Work status was mentioned in 52 conversations. Patients’ or caregivers’ discussions across social media platforms were centered around the symptoms (61.9%, n=287) and causes (58%, n=269) of ocular pain. Patients mentioned having symptoms associated with COSP, including headache or head pressure, dry or gritty eyes, light sensitivity, etc. Patients posted that their COSP impacts day-to-day activities such as reading, driving, sleeping, and their social, mental, and functional well-being. CONCLUSIONS Insights from this study reported patients’ experiences, concerns, and the adverse impact on overall QoL. COSP imposes a significant burden on patients, which spans multiple aspects of daily life.

BACKGROUND Graft-versus-host disease (GVHD) is the major cause of short- and long-term morbidity and mortality after allogeneic hematopoietic stem cell transplantation. Treatment options beyond corticosteroid therapy remain limited, and prolonged treatment often leads to impaired quality of life (QoL). A better understanding of the needs and experiences of patients with GVHD is required to improve patient care. OBJECTIVE The aim of this study is to explore different social media (SM) channels for gathering and analyzing the needs and experiences of patients and other stakeholders across 14 European countries. METHODS We conducted a retrospective analysis of SM data from the public domain. The Talkwalker social analytics tool collected data from open-access forums, blogs, and various social networking sites using predefined search strings. The raw data set derived from the aggregator tool was automatically screened for the relevancy of posts, generating the curated data set that was manually reviewed to identify posts that fell within the predefined inclusion and exclusion criteria. This final data set was then used for the deep-dive analysis. RESULTS A total of 9016 posts relating to GVHD were identified between April 2019 and April 2021. Deduplication and relevancy checks resulted in 325 insightful posts, with Twitter contributing 250 (77%) posts; blogs, 49 (15%) posts; forums, 13 (4%) posts; Facebook, 7 (2%) posts; and Instagram and YouTube, 4 (1%) posts. Patients with GVHD were the primary stakeholders, contributing 63% of all SM posts. In 234 posts, treatment was the most discussed stage of the patient journey (68%), followed by symptoms (33%), and diagnosis and tests (21%). Among treatment-related posts (n=159), steroid therapy was most frequently reported (54/159, 34%). Posts relating to treatment features (n=110) identified efficacy (45/110, 41%), side effects (38/110, 35%), and frequency and dosage (32/110, 29%), as the most frequently discussed features. Symptoms associated with GVHD were described in 24% (77/325) of posts, including skin-related conditions (49/77, 64%), dry eyes or vision change (13/77, 17%), pain and cramps (16/77, 21%), and fatigue or muscle weakness (12/77, 16%). The impacts of GVHD on QoL were discussed in 51% (165/325) of all posts, with the emotional, physical and functional, social, and financial impacts mentioned in 69% (114/165), 50% (82/165), 5% (8/165), and 2% (3/165) of these posts, respectively. Unmet needs were reported by patients or caregivers in 24% (77/325) of analyzed conversations, with treatment-related side effects being the most common (35/77, 45%) among these posts. CONCLUSIONS SM listening is a useful tool to identify medical needs. Treatment of GVHD, including treatment-related side effects, as well as its emotional and physical impact on QoL, are the major topics that GVHD stakeholders mention on SM. We encourage a structured discussion of these topics in interactions between health care providers and patients with GVHD. CLINICALTRIAL Not applicable

Background: Cutaneous melanoma is an aggressive malignancy that is proposed to account for 90% of skin cancer–related mortality. Individuals with melanoma experience both physical and psychological impacts associated with their diagnosis and treatment. Health-related information is being increasingly accessed and shared by stakeholders on social media platforms. Objective: This study aimed to assess how individuals living with melanoma across 14 European countries use social media to discuss their needs and provide their perceptions of the disease. Methods: Social media sources including Twitter, forums, and blogs were searched using predefined search strings of keywords relating to melanoma. Manual and automated relevancy approaches filtered the extracted data for content that provided patient-centric insights. This contextualized data was then mined for insightful concepts around the symptoms, diagnosis, treatment, impacts, and lived experiences of melanoma. Results: A total of 182,400 posts related to melanoma were identified between November 2018 and November 2020. Following exclusion of irrelevant posts and using random sampling methodology, 864 posts were identified as relevant to the study objectives. Of the social media channels included, Twitter was the most commonly used, followed by forums and blogs. Most posts originated from the United Kingdom (n=328, 38%) and Spain (n=138, 16%). Of the relevant posts, 62% (n=536) were categorized as originating from individuals with melanoma. The most frequently discussed melanoma-related topics were treatment (436/792, 55%), diagnosis and tests (261/792, 33%), and remission (190/792, 24%). The majority of treatment discussions were about surgery (292/436, 67%), followed by immunotherapy (52/436, 12%). In total, 255 posts discussed the impacts of melanoma, which included emotional burden (n=179, 70%), physical impacts (n=61, 24%), effects on social life (n=43, 17%), and financial impacts (n=10, 4%). Conclusions: Findings from this study highlight how melanoma stakeholders discuss key concepts associated with the condition on social media, adding to the conceptual model of the patient journey. This social media listening approach is a powerful tool for exploring melanoma stakeholder perspectives, providing insights that can be used to corroborate existing data and inform future studies.

BACKGROUND Cutaneous melanoma is an aggressive malignancy that is proposed to account for 90% of skin cancer–related mortality. Individuals with melanoma experience both physical and psychological impacts associated with their diagnosis and treatment. Health-related information is being increasingly accessed and shared by stakeholders on social media platforms. OBJECTIVE This study aimed to assess how individuals living with melanoma across 14 European countries use social media to discuss their needs and provide their perceptions of the disease. METHODS Social media sources including Twitter, forums, and blogs were searched using predefined search strings of keywords relating to melanoma. Manual and automated relevancy approaches filtered the extracted data for content that provided patient-centric insights. This contextualized data was then mined for insightful concepts around the symptoms, diagnosis, treatment, impacts, and lived experiences of melanoma. RESULTS A total of 182,400 posts related to melanoma were identified between November 2018 and November 2020. Following exclusion of irrelevant posts and using random sampling methodology, 864 posts were identified as relevant to the study objectives. Of the social media channels included, Twitter was the most commonly used, followed by forums and blogs. Most posts originated from the United Kingdom (n=328, 38%) and Spain (n=138, 16%). Of the relevant posts, 62% (n=536) were categorized as originating from individuals with melanoma. The most frequently discussed melanoma-related topics were treatment (436/792, 55%), diagnosis and tests (261/792, 33%), and remission (190/792, 24%). The majority of treatment discussions were about surgery (292/436, 67%), followed by immunotherapy (52/436, 12%). In total, 255 posts discussed the impacts of melanoma, which included emotional burden (n=179, 70%), physical impacts (n=61, 24%), effects on social life (n=43, 17%), and financial impacts (n=10, 4%). CONCLUSIONS Findings from this study highlight how melanoma stakeholders discuss key concepts associated with the condition on social media, adding to the conceptual model of the patient journey. This social media listening approach is a powerful tool for exploring melanoma stakeholder perspectives, providing insights that can be used to corroborate existing data and inform future studies.