Tools & Platforms

Lay beliefs about AI assessment of interpersonal skills in personnel selection

Published

2 months ago

July 13, 2025

Studies 1–3: Lay beliefs about AI assessing interpersonal skills

Studies 1 to 3 find lay beliefs about AI assessing applicants’ interpersonal skills. Study 1 tested which pieces of information people tend to check when asked to guess whether the applicant had passed the interview assessed by either an AI or human, from four assessment scores (including interpersonal and analytical skills assessment scores). Participants (n = 200) could only choose to reveal one of the four assessment scores and were informed that a bonus would be paid for a correct guess (i.e., incentive-compatible design). Individuals tend to rely on and give more weight to information that aligns with their beliefs^43,44. Therefore, if individuals doubt AI’s capabilities to assess interpersonal skills, they may rely on what they perceive as more reliable information. Consequently, participants in the AI (versus human) condition would be less likely to seek interpersonal skills assessment scores than analytical skills assessment scores.

As pre-registered, we compared the proportion of participants who chose to view the interpersonal skills assessment score versus the analytical skill assessment score as a function of assessment agent type. As predicted, the assessment agent type significantly influenced the decision of which information to view (χ²(1) = 19.28, P < 0.001). In the AI condition, 42.7% of participants chose the interpersonal skills score, while 75.3% did in the human condition. The odds of choosing to view interpersonal skills over analytical skills assessment scores were 4.09 times higher (95% CI[2.15, 7.77]) in the human than in the AI condition.

Study 2 tested lay beliefs through another experiment. Participants (n = 201) were randomly assigned to one of two selection processes assessed by either AI or humans. Participants in the AI [Human] condition read a brief description of AI interviewers [HR personnel] and rated the capabilities of the AI interviewers [HR personnel] in assessing job candidates’ interpersonal and analytical skills, respectively (e.g., AI interviewers would be capable of evaluating the job candidates’ interpersonal skills.; 7-point Likert scale; 7 = strongly agree). We expected participants to show less trust towards AI (vs. humans) in assessing interpersonal skills but not analytical skills.

We examined the perceived capabilities of the AI versus human (i.e., assessment agent; between-participants factor) in assessing interpersonal and analytical skills (i.e., skill type; within-participants factor) using a mixed ANOVA. Our hypothesis was supported by the significant interaction effect (F(1, 197) = 93.86, P < 0.001, η² = 0.323), showing a lower perceived capability of AI (vs. human) for evaluating interpersonal skills but not analytical skills (Fig. 1). Specifically, planned contrast revealed that participants perceived the AI as less capable in the assessment of interpersonal skills (M = 3.28, SD = 1.77) compared to humans (M = 5.24, SD = 1.30; F(1, 197) = 80.07, P < 0.001, η² = 0.289). However, participants perceived that AI’s capabilities in assessing analytical skills (M = 4.86, SD = 1.60) were comparable to that of humans (M = 4.64, SD = 1.41; F(1, 197) = 1.10, P = 0.30, η² = 0.006).

Fig. 1

Perceived capabilities as a function of selection agent and skill type in Study 2. Note: Error bars indicate 1 standard error of the mean.

We replicated Study 2 (n = 63) using a similar design but replacing the term AI interviewer with AI assessment tools designed to analyze candidates’ responses to job- and personality-related questions and their performance in mini-games^2,3. The results were consistent with our earlier findings (see SI 5 for details).

Study 3 (n = 343) explored lay beliefs about AI selection with practitioners by directly asking managers to compare their perceptions about AI and human assessments. We used surveys collected by MidasIT, a major provider of AI selection technology in Korea. Of the participants, 81.6% (n = 280) were responsible for hiring tasks. We focused on two questions asking whether they believed AI or humans were better at predicting an applicant’s interpersonal and analytical skills, respectively (1⁷ indicated definitely AI [human] interviews). We expected managers to believe AI is less effective than humans in predicting an applicant’s interpersonal skills but not analytical skills.

As predicted, managers rated AI as less capable than humans in predicting interpersonal skills (M = 5.02, SD = 1.37; one-sample t-test comparing against 4: t(342) = 13.81, P < 0.001, Cohen’s d (d) = 0.75, 95% CI[0.88, 1.17]), but as more capable in predicting analytical skills (M = 3.44, SD = 1.61; one-sample t-test: t(342) = −6.47, P < 0.001, d = −0.35, 95% CI[−0.73, −0.39]). Next, we estimated a linear mixed model with a random intercept and slope for each manager, where we regressed managers’ lay beliefs on each of the two assessed skill types while controlling for their involvement in hiring tasks (i.e., by adding fixed effects for a dummy variable indicating whether each respondent was in charge of hiring tasks or not). The results showed the same pattern of significance (b = 1.58, t(342) = 14.88, P < 0.001, 95% CI[1.37, 1.58]).

Studies 4–5: Consequences of lay beliefs about AI assessing interpersonal skills

The previous studies suggest people negatively perceive AI’s capabilities in assessing applicants’ interpersonal (vs. analytical) skills. Studies 4 and 5 examine the potential consequences of these lay beliefs for office workers and job applicants. Study 4 tested whether office workers hold biases towards AI-selected employees as opposed to those selected by humans. If office workers are skeptical of AI’s capabilities in assessing interpersonal skills, they might also perceive AI-selected employees as having inferior interpersonal skills than those selected by humans. As a result, we predict that office workers would be less likely to assign tasks requiring interpersonal skills to AI-selected employees (vs. human-selected), but this bias would not extend to tasks requiring analytical skills. That is, we predict that office workers would not assign tasks requiring analytical skills more to employees selected by humans than AI.

We recruited 102 office workers in large Korean companies, which are increasingly using AI assessments. From the managers’ perspective, the office workers evaluated two hypothetical employees, one hired through a conventional HR interview and the other hired through an AI interview. Participants were then asked to indicate which employee they would assign tasks requiring interpersonal skills and analytical skills, respectively (i.e., using a 7-point Likert scale where 1⁷ indicated a definite preference for the HR [AI] interview hired employee).

As predicted, the office workers indicated that they would assign tasks requiring interpersonal skills less to the employee hired through the AI interview than the one hired through the human interview (M = 2.55, SD = 1.36; one sample t-test with a comparison against 4: t(101) = −10.76, P < 0.001, d = −1.07, 95% CI[−1.72, −1.18]). However, they were more likely to assign tasks requiring analytical skills to the AI-interviewed employee (M = 5.29, SD = 1.41; one sample t-test: t(101) = 9.26, P < 0.001, d = 0.92, 95% CI[1.02, 1.57]).

Study 5 tested another potential consequence of the lay beliefs, focusing on applicants’ strategies during the selection process assessed by AI versus humans. We investigated which type of skills is considered important for each of the two selection types. Given that people’s behavior tends to reflect their beliefs^43,44, if job applicants are skeptical about AI’s capabilities in assessing interpersonal skills, applicants may infer that showcasing these skills will have less impact on the selection decision. Consequently, they should emphasize their interpersonal skills less in the AI-assessed selection process than in the human-assessed one. Conversely, we anticipated no such difference for analytical skills.

We recruited 107 participants from South Korea, who had experienced a selection process evaluated by both AI and humans. We asked participants to recall their most recent interviews of each type and then to indicate in which of the two interviews they had emphasized their interpersonal and analytical skills relatively more (where 1⁷ indicated they definitely emphasized the skill more in the human [AI] interview; the order of the two skill questions was counterbalanced, which has no significant effects on participants’ responses (all p-values > 0.4; SI 6)).

As predicted, participants emphasized interpersonal skills less during the AI interview than human interview (M = 3.31, SD = 1.67; one sample t-test against 4: t(106) = −4.28, P < 0.001, d = −0.41, 95% CI[−1.01, −0.37]). On the other hand, participants emphasized analytical skills more during the AI interview than human interview (M = 4.42, SD = 1.51; one sample t-test against 4: t(106) = 2.88, P = 0.01, d = 0.28, 95% CI[0.13, 0.71]).

Study 6: Alleviating negative lay beliefs about AI assessing interpersonal skills

The previous studies highlight negative lay beliefs about AI that could cause problems in organizational settings, affecting managers’ task assignments and job applicants’ strategies during selection processes. However, our field data analysis suggests that AI selection technology can outperform humans in predicting interpersonal performances. Likewise, research continues to demonstrate the efficacy of AI in measuring interpersonal traits, such as friendliness^26,27,28,29. Therefore, Study 6 investigated a strategy to reduce the negative lay beliefs about AI assessing interpersonal skills. We speculated that people may find it challenging to grasp the concepts of AI technologies being used to assess interpersonal skills. Therefore, we predicted that informing people of various examples of recently developed AI selection technologies that can assess interpersonal skills would help reduce the negative lay beliefs.

To test this, we recruited 231 participants who had heard of AI selection technologies to examine the change in consumers’ existing beliefs toward AI’s capabilities. We told participants that they would take on two supposedly unrelated tasks. In the first task, participants were randomly assigned to one of the two conditions—AI-information vs. control condition. Participants in the AI-information condition summarized a news article about the trend of AI selection technologies that capture facial expressions, emotions, social skills, and more, while those in the control condition summarized a news article on the use of AI in separate contexts. For the second task, participants were asked to evaluate which type of the two selection processes would be more effective in selecting applicants with outstanding interpersonal and analytical skills, respectively (where 1⁷ represents definitely more effective in human [AI] interview).

As in Study 2, participants in the control condition thought that AI was less effective than humans in selecting applicants with outstanding interpersonal skills (M = 2.63, SD = 1.56; one sample t-test: t(114) = −9.42, P < 0.001, d = −0.88, 95% CI[−1.66, −1.08]); however, they thought AI was more effective for selecting applicants with outstanding analytical skills (M = 5.16, SD = 1.47; t(114) = 8.46, P < 0.001, d = 0.79, 95% CI[0.89, 1.43]). Using a mixed ANOVA, we then tested whether informing participants about advanced AI selection technologies would increase their perception of AI (vs. humans) in assessing interpersonal skills, but not in terms of analytical skills (Fig. 2). As expected, the interaction was significant (F(1, 229) = 7.67, P = 0.01, η² = 0.032). A further breakdown of this interaction showed that the perceived capabilities of AI assessing interpersonal skills were higher in the AI-information condition (M = 3.28, SD = 1.67) than in the control condition (M = 2.63, SD = 1.56; F(1, 229) = 9.34, P = 0.03, η² = 0.039). By contrast, the perceived effectiveness of AI in assessing analytical skills did not differ significantly between the two conditions (M_{AI−information} = 5.16, SD = 1.47 vs. M_control = 5.08, SD = 1.32; F(1, 229) = 0.19, P = 0.67, η² = 0.01).

Source link

Up Next

As the AI jobs armageddon approaches, it seems that only …

Don't Miss

AI opens new worlds for the blind

Young Eun Huh

Click to comment

Tools & Platforms

Laptop or tablet? Plus, power packs, headphones and AI advice – The Irish Times

Published

52 minutes ago

September 11, 2025

Ciara O'Brien

College students are coming back to their lecture halls this week or starting out on their third-level studies. If you are an incoming first year, there is plenty to get to grips with, from reading lists and schedules to travel passes and finding your way around the campus.

On top of that, you probably have some extra technology requirements. Perhaps a laptop upgrade is on the cards or you might think of investing in a tablet.

So what are the tech essentials and what are simply nice to have?

Laptops

The laptop is not yet dead, despite some predictions of its demise. In fact, things seem to be going well for the device the tablet was supposed to replace.

Picking the right one can be difficult, though; you need to strike a balance between a laptop that will last a few years and not overpaying for technology you don’t actually need. In other words, the top-spec machines are great but they aren’t necessary for everyone.

There are some things to consider before you make your choice. First is the size of the device. If you plan on carrying it around everywhere, something reasonably compact and light is best. Most people find the 13- to 14-inch screens offer the best balance in terms of portability while still giving you enough room to work with.

That should cover you for creating documents, browsing online, editing images and videos, and creating content for your social channels if that is what you are into.

For those who need something with a bit more screen space – for design, perhaps – a 15-inch display might suit better, but will be more clunky to carry around.

Once you have decided on the size, it is time to look at its main components. Think of the processor as the brain of the computer. On a budget, the AMD Ryzen 3 series or Intel Core i3 will be fine for web browsing and word processing.

Moving up the chain, AMD’s Ryzen 5 offers a good balance of performance and efficiency, comparable to the Intel Core i5 chip. For more demanding tasks, the Ryzen 7 or Core i7 chips are worth considering, although they will incur a higher cost.

Then you get into RAM, which will allow your computer to deal with heavier workloads and run more tasks at once without stuttering to a halt. Windows requires a minimum of 4GB. Bumping it to 8GB is sufficient for basic workloads.

To ensure your computer lasts longer, consider upgrading to 16GB if possible. Video editing, photo editing, gaming and other power-hungry applications will need more RAM, so consider your future needs too.

And when it comes to on-board storage, aim for 512GB if you can. You can supplement it with external and cloud storage, but that requires either carrying around a second device – not ideal if you want to stay mobile – or an internet connection, which isn’t always possible.

Mac versus Windows is the ongoing battle, but ChromeOS is also an option for students on a budget, or Linux if you are technically inclined. The decision will ultimately depend on what software you need to run. If you have no specific requirements, then your choices are much wider. Some students will need to run specific software for their course – computer-aided design programmes, for example – that will have minimum system requirements.

Power

If you have a new-ish laptop with a power-efficient chip, you probably won’t run into too many issues with your device’s battery. But occasionally, they take a battering, and when you find your battery dwindling to alarming levels, having some sort of backup is handy.

Doing more with less is a good approach, and these days most devices can be charged over USB-C, thanks to some heavy nudges from the European Commission. In theory, you could carry one charger in your bag and power all your devices at various times throughout the day.

If you need more than one device charged, though, something with multiple USB options is a good idea. Take the Anker GaNPrime 200W Charger (€90, Harvey Norman). Not only does it charge both USB-A and USB-C devices, it can also power up to six devices simultaneously.

Anker GaNPrime 200W charger

It has a total power rating of 200W, which means you can charge two 14-inch MacBook Pro devices in under half an hour or several USB-C-enabled smartphones simultaneously. It also has a built-in protection system to make sure your devices charge safely and at a suitable temperature.

However, there will inevitably be a time when you are away from a power outlet and caught with a low battery. The Chargeasap Flash Pro Plus ($200, chargeasap.com), a graphene 25000mAh battery, cannot only charge your smartphone and Apple Watch, but also your laptop.

Noise-cancelling headphones

If you don’t already own a pair of these, you need to invest in some, asap. Not only can they help you to block out the bus noise on the commute home, they may also save your sanity in the library around exam time. Among the best in the over-ear category are the Sony WH-1000XM6 (€450, Currys).

Apple users may prefer the in-ear AirPods Pro (€279, Apple), but there are plenty of options out there that will cross the platform barrier and work with your laptop too, such as the Bose QuietComfort Active ear buds (€150, Harvey Norman)

Apple AirPod 2 with USB Type C charing port. Photograph: Ming Yeung/Getty

If you are just in need of some decent earplugs without any added distraction, however, the Loop Quiet 2 (€20, Loop.eu) will block out enough noise to help you focus.

AI

Artificial intelligence is creeping into our daily lives, from AI-enabled chatbots talking us through customer service nightmares to research assistants that can create podcasts from our documents and debate the finer points of our research topic.

The capabilities out there vary, however. Apple is still trailing slightly on AI integration. although devices such as the iPhone 15 Pro and Pro Max, and newer models, support the handful of Apple Intelligence features that have been launched in Europe. That includes tools to manage your inbox, refine messages and documents, and edit photographs. Broader searches call in ChatGPT.

Samsung and Google have taken it a step further. Gemini, Google’s AI-enabled assistant, has now become the default assistant on your smartphone, with its latest Gemini machine learning models and chatbots integrated into the system.

There are other ways to use AI, however, assuming your college course allows it. Grammarly, for example, will help you polish up college projects and essays, while also keeping an eye out for potential plagiarism, assuming you take out the Pro subscription at €144 for the year. Its verdict on this article, for example, was a resounding “zero copied text”.

[ I have advice for prospective third-level students, and it’s not what you thinkOpens in new window ]

ChatGPT, meanwhile, can help you research a specific topic with its latest GPT-5 model which, according to founder Sam Altman, has “PhD level intelligence”. Whether you buy into that or not, ChatGPT can be a handy research partner. Just remember, as with all AI, to check up on its sources before you believe what it tells you.

The current favourite AI-enabled tool around here though is Google’s NotebookLM. Free to use, you can create notebooks for research topics with up to 50 sources each. The AI will then analyse the sources and explain it to you in bullet points, create an FAQ-style sheet or study guide, or even generate an AI podcast, where two synthetic hosts discuss the topic in detail.

Source link

Tools & Platforms

Deutsche Telekom Leverages n8n’s Automation Platform to Power Next-Gen AI Agents

Published

1 hour ago

September 11, 2025

Ray Sharma

AI agents are currently the dominant topic in the digitization scene. Deutsche Telekom announced its partnership with n8n. The German startup from Berlin, which is popular with developers and investors, is a specialist in the digitization of repetitive workflows.

n8n is currently valued far beyond the threshold of 1 billion euros. Deutsche Telekom relies on n8n for the development of AI agents for its business customer division. Whether in production, logistics, accounting or sales and marketing: the application scenarios are diverse, especially in medium-sized companies.

AI agents for all cases

AI agents are software programs that independently plan tasks, take measures and make decisions for a specific goal with the help of artificial intelligence, without constant human guidance. N8n is an open-source workflow automation platform. Depending on the use case, different apps, interfaces and services can be combined with each other. The code is created by the AI. Only basic programming knowledge is required. The offer is primarily aimed at medium-sized companies.

Three different models serve as the basis for the development of tailor-made AI agents:

Out-of-the-box standard AI agents for simple tasks such as appointment finders or interactive voice response systems (IVR)

AI agent modular system for creating custom agent flows via drag and drop

Fully individualized AI agent flows with customized service

The first pilot projects are testing its use, for example in logistics: Until now, freight documents have been passed on manually. This process is now taken over by an AI agent.

Deutsche Telekom has been using artificial intelligence for many years. The group relies on partnerships for the best AI solutions on the market and collaborative development. The following applies: AI should solve business problems and increase productivity – while maintaining a focus on the well-being of people.

Gartner predicts that by 2028, AI agents will take over 20 percent of interactions in online stores. In the workplace, they will make 15 percent of everyday decisions, according to the market research institute, up from zero percent in 2024. And: By 2028, 33 percent of software in companies will contain agentic AI, compared to less than one percent in 2024.

Maximilian Ahrens, Managing Director T Digital, Telekom Deutschland

We are very proud that we are working with what is currently the most prominent tech startup from Germany. With n8n’s platform and technology, we make it easy for our customers to implement AI-supported processes and future-proof their business

Source link

Tools & Platforms

Shandong Inspur Intelligent Building Technology Showcases AI-Driven Solutions at BEX Asia 2025

Published

3 hours ago

September 11, 2025

The Editors

SINGAPORE, Sept. 11, 2025 /PRNewswire/ — Shandong Inspur Intelligent Building Technology Co., Ltd. (“Inspur Intelligent Building”), a leading provider of smart building solutions, took center stage at BEX Asia 2025, unveiling its latest advancements in AI-powered smart building systems. Organized by leading global event management company Reed Exhibitions, the exhibition opened on September 3, 2025, at the Marina Bay Sands Expo & Convention Centre in Singapore.

As one of the largest specialized building and construction exhibitions in Southeast Asia, this year’s edition introduced an innovative “four-in-one” exhibition format, co-locating InnoBuild, MCE Asia, and Smart Cities & Building Asia. Centered around the core themes of “Productivity, Sustainability, Energy Efficiency, and Intelligence”, the integrated events brought together resources from across the construction ecosystem to build a professional platform for technology exchange, trend exploration and partnership development.

BEX Asia 2025 attracted over 3,000 global leaders, entrepreneurs, and policymakers across more than 150 exhibition booths. During the event, industry frontrunners expressed strong interest in fostering dialogue on innovation and accelerating the transition to a more sustainable and connected built environment. As a leader in smart building transformation, Inspur Intelligent Building showcased its latest breakthroughs and real-world applications in the integration of artificial intelligence and building technologies, offering scalable, future-ready solutions for developers, owners, and operators.

Exhibition booth at ASEAN AI Summit

At the exhibition, Inspur Intelligent Building showcased its integrated applications of AI across the building lifecycle:

Modular Building Systems: Combining efficient prefabrication, flexible assembly, and intelligent adaptation to deliver cost-effective residential and commercial spaces, reinforcing competitiveness in global markets and the company’s commitment to sustainable building practices;
AI-Enhanced Design Automation: Enabling human-machine collaboration to standardize and streamline construction design processes;
Intelligent Site Management Platform: Incorporating AI-powered image recognition, advanced query tools, building information modeling (BIM), and point-cloud imaging to enable real-time project monitoring and precision management of project resources;
Smart Living Solutions: Delivering a unified IoT platform for whole-home connectivity and intelligent automation, enhanced by large AI models to create personalized occupant experiences;
Sustainable Building Technologies: Offering a cloud-edge product suite including AI-driven energy optimization, energy and carbon management platforms, and AI edge workstations, designed to help clients reduce energy use and carbon footprint.

On-site consultation with a customer

Throughout the exhibition, visitors from Singapore, Europe, Southeast Asia and other global markets engaged with Inspur Intelligent Building’s technical team to explore its comprehensive portfolio for AI-assisted design, green building systems, energy and carbon management platforms, smart campus and industrial park systems, intelligent construction sites, and prefabricated steel structures. Live demonstrations and case studies illustrated the company’s engineering expertise and real-world applications, earning positive recognition from industry partners.

This first appearance at BEX Asia 2025 marked an important milestone in Inspur Intelligent Building’s international expansion and opened new growth opportunities in Southeast Asia. Looking ahead, the company plans to continue investing in R&D for AI-enabled smart building technologies, grow its solution ecosystem, and accelerate international growth—supporting the digital, intelligent, and sustainable transformation of the built environment worldwide.

Source: Shandong Inspur Intelligent Building Technology Co., LTD.

Source link