AI and Digital Literacy: Impact on Information Resilience in Indonesian Society

Abstract

This study examines the relationship between artificial intelligence (AI) media narratives and digital literacy levels in Indonesia, focusing on their combined impact on information resilience and misinformation vulnerability. Through a mixed-methods approach involving content analysis of five major Indonesian media outlets (2022–2024), a survey of 384 respondents across South Sulawesi, and in-depth interviews with 15 media professionals, this research reveals significant gaps between AI media coverage and public understanding. The findings indicate that while 76% of Indonesian media coverage presents AI as a technological solution to misinformation, only 32% of the surveyed population demonstrates adequate digital literacy skills to evaluate AI-generated content critically. The study identifies three distinct patterns of AI media framing: technological optimism (47%), cautionary realism (33%), and dystopian skepticism (20%). These findings contribute to understanding how media narratives about emerging technologies influence public digital literacy and information resilience, particularly in developing digital economies where technological adoption often outpaces digital education initiatives.

1. Introduction

The proliferation of artificial intelligence technologies in Indonesian media landscapes presents unprecedented opportunities and complex challenges for public information resilience. As Indonesia continues its digital transformation journey, with internet penetration reaching 79.5% by 2024 according to the Indonesian Internet Service Providers Association (Santika, 2024), the intersection of AI adoption and digital literacy becomes increasingly critical for maintaining information integrity (Sutrisno, 2024). The emergence of generative AI technologies has fundamentally altered how information is created, disseminated, and consumed, creating new vulnerabilities in an already complex information ecosystem.

Recent developments in Indonesia’s media landscape demonstrate the urgency of understanding these dynamics. The country has experienced significant challenges with misinformation and disinformation, particularly during major political events and health crises (Lim, 2017). The COVID-19 pandemic further highlighted the vulnerability of Indonesian society to false information, with various studies documenting widespread circulation of health-related misinformation across digital platforms (Juditha, 2018; Nasir et al., 2021). Against this backdrop, introducing AI technologies presents a paradox where the technologies promoted as solutions to misinformation challenges may inadvertently create new forms of information vulnerability.

The Indonesian media’s approach to covering AI technologies reflects broader global patterns while maintaining distinct local characteristics. Indonesian journalism has traditionally played a crucial role in shaping public understanding of technological innovations, often serving as the general population’s primary source of information about complex technical subjects (Sen & Hill, 2000). However, rapid AI development has challenged traditional journalistic practices, requiring reporters to navigate unfamiliar technical territories while maintaining accuracy and accessibility in their coverage.

Digital literacy levels in Indonesia present a complex picture that varies significantly across demographic and geographic lines. While urban areas, particularly in Java, demonstrate higher digital literacy rates, rural regions and older populations face substantial challenges navigating digital information environments (Kurnia & Astuti, 2017). This digital divide becomes particularly problematic when considering AI technologies, which require a sophisticated understanding of algorithmic processes, data sources, and potential biases to evaluate critically.

Information resilience has gained prominence in media studies as researchers seek to understand how societies can maintain information integrity in increasingly complex digital environments (Humprecht et al., 2020). Information resilience encompasses the ability to identify false information and the capacity to adapt to new forms of information manipulation and maintain trust in legitimate information sources. In AI technologies, information resilience requires understanding how these systems work, their limitations, and their potential for beneficial and harmful applications.

This study draws primarily on media framing theory (Entman, 1993), which provides insights into how journalistic choices shape public understanding of complex issues. The combination of framing analysis with digital literacy measurement represents a novel methodological approach examining media production and audience reception processes.

Diakopoulos (2019) notes that “The increasing reliance on algorithmic systems for information processing and dissemination creates new challenges for maintaining transparency and accountability in news production”. This observation becomes particularly relevant in the Indonesian context, where rapid AI adoption in media organizations has often proceeded without corresponding investments in public education or professional training (Newman, 2024).

This study addresses a critical gap in understanding how media narratives about AI technologies influence Indonesia’s public digital literacy and information resilience. While previous research has examined either AI adoption patterns or digital literacy levels independently, limited attention has been paid to their interconnected dynamics and combined impact on information vulnerability. The Indonesian context provides a particularly valuable case study due to the country’s position as a rapidly developing digital economy with diverse linguistic, cultural, and educational backgrounds.

Three primary research questions guide this study: How do Indonesian media outlets frame artificial intelligence technologies about information challenges and solutions? Second, what is the relationship between exposure to AI media narratives and individual digital literacy levels among Indonesian internet users? Third, how do these factors influence information resilience and vulnerability to misinformation in different demographic segments of Indonesian society? The main objective is to provide empirical evidence for developing evidence-based policies and educational interventions to enhance information resilience in AI-mediated environments.

2. Literature Review

2.1. AI in Media and Journalism: Global Perspectives

The integration of artificial intelligence in journalism has emerged as a transformative force, reshaping news production, distribution, and consumption globally. Diakopoulos (2019) provides a comprehensive framework for understanding algorithmic journalism, highlighting how automated systems increasingly influence editorial decisions, content creation, and audience engagement strategies. This transformation extends beyond simple automation to encompass complex information curation, fact-checking, and even narrative construction processes.

Recent systematic reviews of AI in journalism reveal both opportunities and challenges associated with technological adoption in newsrooms. The research Sonni et al. (2024, p. 1565) demonstrates that 73% of news organizations globally have adopted AI technology, with significantly varying implementation approaches and outcomes. Their findings indicate that “while AI technologies can enhance efficiency and accuracy in certain contexts, they also introduce new forms of bias and dependency that require careful management and oversight”.

The Reuters Institute’s annual reports on journalism and technology trends consistently highlight the growing influence of AI systems on news production and consumption patterns (Newman, 2024). These reports document how AI technologies are reshaping how news is produced and how audiences discover, consume, and engage with information. Newman (2024, p. 18) Notes that “the emergence of AI-powered recommendation systems, automated content generation, and sophisticated personalization algorithms has created new dynamics in the relationship between news organizations and their audiences”.

European research on AI in journalism focuses on ethical implications and regulatory challenges. Helberger (2019, p. 995) Examines the democratic consequences of algorithmic news curation, arguing that “AI systems in journalism carry significant responsibilities for maintaining diverse and balanced information environments”. This perspective has influenced policy discussions within the European Union and other regulatory frameworks seeking to balance innovation with democratic values.

2.2. Media Framing of Technology and Innovation

Media framing theory provides essential insights into how journalistic choices shape public understanding of complex technologies. The seminal work Entman (1993) demonstrates how framing processes influence what audiences think about and how they think about specific issues. In the context of AI technologies, framing choices become particularly significant due to these systems’ technical complexity and potential societal implications.

Research on technology framing in news media reveals consistent patterns across different innovations and national contexts. Nisbet and Scheufele (2009) Identify standard science and technology coverage frames, including progress narratives, risk assessments, and ethical concerns. These frames significantly influence public attitudes toward technological adoption and regulation.

Studies focusing on AI coverage in news media demonstrate the prevalence of utopian and dystopian narratives, often with limited middle-ground perspectives (Brennen & Nielsen, 2018). This polarized coverage can contribute to public confusion and polarization regarding AI technologies and their appropriate regulation and implementation. Brennen and Nielsen (2018, p. 1) argue that “news coverage of artificial intelligence tends toward sensationalism, presenting either revolutionary benefits or catastrophic risks with little nuanced analysis”.

2.3. Digital Literacy and Information Resilience

Digital literacy has evolved from basic computer skills to encompass complex competencies for navigating contemporary information environments. The framework Buckingham (2007) emphasizes critical evaluation skills, understanding digital media production processes, and awareness of commercial and political influences on digital content. These competencies become particularly crucial when considering AI-generated or AI-mediated content, which may not follow traditional indicators of credibility and reliability.

Research on information resilience has gained prominence as scholars seek to understand how individuals and communities maintain information integrity in challenging environments. Humprecht et al. (2020, p. 501) conceptualize information resilience encompassing multiple dimensions, including media literacy, institutional trust, and social cohesion. Their comparative analysis across different national contexts reveals significant variations in resilience factors and their relative importance. They note that “information resilience is not simply a matter of individual competencies but requires supportive institutional and social environments”.

The concept of algorithmic awareness has emerged as a crucial component of digital literacy in AI-mediated environments. Eslami et al. (2015) Demonstrate that many internet users lack a basic understanding of how algorithmic systems shape their information experiences, leading to decreased ability to evaluate information sources and potential biases critically. This research has informed educational initiatives to improve public understanding of algorithmic processes and their implications.

Studies on developing countries reveal additional challenges in building digital literacy and information resilience. Oyedemi (2012, p. 308) Examines digital inequality in Africa, highlighting how “economic, educational, and infrastructural constraints limit effective digital participation”. Similar patterns have been observed in Southeast Asia, where rapid technological adoption often outpaces educational adaptation and regulatory development.

Recent developments in digital literacy research emphasize the importance of AI-specific competencies. UNESCO (2023) Demonstrates that traditional digital literacy approaches require updating to address algorithmic decision-making and automated content generation. The European Union’s DigComp 2.2 framework (Vuorikari et al., 2022) similarly highlights the emergence of “AI awareness” as a distinct competency requiring specialized educational approaches, particularly in understanding how algorithmic systems shape information experiences.

Furthermore, research Pangrazio and Selwyn (2018) emphasizes the critical importance of “personal data literacies” in AI-mediated environments, arguing that citizens must understand how to use digital technologies and how these technologies collect, process, and utilize personal information for algorithmic decision-making.

2.4. Indonesian Media Landscape and Digital Transformation

Indonesia’s media landscape has undergone a significant transformation since the fall of the Suharto regime in 1998, with democratization processes enabling greater media diversity and freedom (Sen & Hill, 2000). However, this transformation has also introduced new challenges related to information quality, media ownership concentration, and regulatory adaptation to digital technologies.

The rise in digital platforms in Indonesia has fundamentally altered media consumption patterns and information distribution mechanisms. Research Lim (2017, p. 415) documents how “social media platforms have become primary news sources for many Indonesians, particularly younger demographics, while traditional media maintain influence among older populations”. This shift has created complex dynamics where different population segments operate within distinct information ecosystems.

Indonesian government responses to digital transformation and information challenges have evolved through various policy initiatives and regulatory frameworks. The Ministry of Communication and Information Technology (Kominfo) has implemented multiple programs to improve digital literacy and combat misinformation, though evaluations of their effectiveness remain limited (Juditha, 2018).

Recent research Surjatmodjo et al. (2024, p. 9) provides crucial insights into the Indonesian information environment, finding that “disinformation spreads six times faster than accurate information on digital platforms”. This finding underscores the urgency of developing robust information resilience capabilities among Indonesian internet users.

3. Methodology

3.1. Research Design

This study employs a convergent mixed-methods design combining quantitative content analysis, survey research, and qualitative interviews to provide a comprehensive understanding of the relationship between AI media narratives and digital literacy in Indonesia. The research design follows Creswell and Plano Clark (2018) recommendations for integrating multiple data sources to address complex social phenomena.

The temporal scope of the study covers the period from January 2022 to December 2024, capturing the rapid evolution of AI technologies and their media coverage during a crucial period of public awareness and adoption. This timeframe encompasses the emergence of mainstream generative AI applications, significant policy discussions, and evolving public discourse about AI implications.

The geographical scope focuses primarily on South Sulawesi province, with Makassar as the primary research site, while incorporating comparative elements from other major Indonesian cities. South Sulawesi was selected for several reasons: its demographic diversity representing urban and rural populations, its position as a significant economic hub outside Java, and practical considerations including research infrastructure and accessibility. This approach allows for detailed local analysis while maintaining broader national relevance through comparative elements and secondary data sources.

3.2. Content Analysis

The selected outlets include Kompas (national newspaper with highest circulation and credibility ratings), Detik.com (leading online news portal with most extensive digital readership), Metro TV (commercial television with strong technology coverage), TVOne (commercial television representing alternative editorial perspectives), and Republika (national newspaper with Islamic orientation, representing religious community perspectives). These outlets were selected based on Nielsen audience ratings, representing approximately 65% of the Indonesian news media market share, and their consistent coverage of technology topics during the study period.

The sampling framework identifies AI-related articles through keyword searches in both Indonesian and English languages, including terms such as “kecerdasan buatan,” “artificial intelligence,” “AI,” “teknologi pintar,” and related variations. A total of 847 articles were analyzed across the study period.

For television content from Metro TV and TVOne, articles were identified through their official websites’ archives and transcript databases.

Table 1. Content analysis sample distribution.

Media Outlet	Article Count	Percentage	Media Type
Kompas	198	23.4%	National Newspaper
Detik.com	234	27.6%	Online Portal
Metro TV	147	17.4%	Television
TVOne	132	15.6%	Television
Republika	136	16.1%	National Newspaper
Total	847	100%

shows the distribution of analyzed articles across different media outlets, representing comprehensive coverage of Indonesia’s major news sources. Both networks maintain searchable digital archives of their news programs. Additionally, the Indonesian Press Council provided media monitoring services to ensure comprehensive coverage of AI-related television content during the study period.

Coding categories examine multiple dimensions of AI coverage, including framing approaches, source utilization, technical accuracy, risk assessment, and policy discussion. Reliability testing achieved intercoder reliability coefficients of 0.87 using Krippendorff’s alpha, indicating strong consistency in coding applications.

Coding categories were developed through a combination of deductive and inductive approaches. Technical accuracy was assessed based on factual correctness of AI capabilities, limitations, and operational mechanisms as verified against academic literature and expert consensus. Source utilization examined the diversity and credibility of quoted sources, categorizing them by institutional affiliation and expertise. Risk assessment coding identified whether articles discussed potential negative consequences, limitations, or challenges associated with AI implementation. Policy discussion coding captured references to regulatory frameworks, governance mechanisms, or policy recommendations. These categories were refined through pilot coding of 50 articles and validated through expert consultation with technology journalism researchers.

3.3. Survey Research

The survey component involves 384 respondents selected through stratified random sampling across South Sulawesi province. The sample size was calculated using Yamane’s formula with a confidence level of 95% and a margin of error of 5%, which is considered adequate for inferential statistics in social science research (Sekaran & Bougie, 2016).

Table 2. Survey sample demographics (n = 384).

Demographic Variable	Category	Frequency	Percentage
Age Group	18–25	134	34.9%
	26–35	115	29.9%
	36–45	87	22.7%
	46+	48	12.5%
Education	Elementary	45	11.7%
	High School	178	46.4%
	University	161	41.9%
Location	Urban	231	60.2%
	Rural	153	39.8%
Gender	Male	192	50.0%
	Female	192	50.0%

presents the demographic characteristics of the 384 survey respondents across South Sulawesi province.

The survey instrument incorporates validated scales for digital literacy assessment, adapting the Digital Media Literacy Scale (Buckingham, 2007) and the Algorithmic Awareness Scale (Eslami et al., 2015) for the Indonesian context and language. Digital literacy measurement encompasses three main dimensions: (1) Technical skills including basic computer operation, internet navigation, and digital tool usage. (2) Critical evaluation abilities covering source credibility assessment, information verification techniques, and bias recognition. (3) Understanding of algorithmic processes relevant to AI-mediated information environments, including knowledge of how recommendation systems work, data collection practices, and automated content generation. Each dimension was measured using 5-point Likert scales with culturally appropriate examples and terminology.

3.4. Qualitative Interviews

The qualitative component includes 15 in-depth interviews with media professionals, digital literacy educators, and technology policy experts. Interview participants were selected through purposive sampling to ensure representation of different perspectives and expertise areas relevant to the research questions.

Table 3. Interview participant categories.

Participant Category	Number	Percentage
Journalists (Technology Beat)	6	40.0%
Digital Literacy Educators	4	26.7%
Technology Policy Experts	3	20.0%
Media Executives	2	13.3%
Total	15	100%

outlines the distribution of interview participants across different professional categories.

Interviews were conducted in Indonesian, recorded with participant consent, and transcribed for analysis. Thematic analysis follows established procedures for identifying, coding, and interpreting patterns across interview data.

Thematic analysis followed Braun and Clarke (2006) six-phase approach: (1) data familiarization through multiple transcript readings; (2) initial code generation using both deductive codes derived from research questions and inductive codes emerging from data patterns; (3) theme identification through systematic pattern recognition across interviews; (4) theme review and refinement through iterative analysis; (5) theme definition and labeling with clear scope boundaries; and (6) final analysis integrating themes with quantitative findings. Two researchers independently coded 20% of transcripts, achieving Cohen’s kappa of 0.83 for intercoder reliability. Themes were validated through regular team meetings and member checking with selected participants.

4. Results

4.1. Convergent Findings: Media Framing and Public Understanding Patterns

Triangulation of content analysis (847 articles), survey data (384 respondents), and expert interviews (15 participants) reveals systematic relationships between media coverage patterns and public AI understanding. As illustrated in

Table 4. Convergent analysis: media framing and public response patterns.

Framing Type	Media Coverage (%)	Public Confidence Score	AI Risk Awareness	Sample Size	Qualitative Validation
Technological Optimism	47.2%	3.84	2.31	147	Theme 2: Editorial preferences (11/15 participants)
Cautionary Realism	32.8%	2.67	3.72	129	Theme 1: Professional challenges (13/15 participants)
Dystopian Skepticism	20.0%	2.15	3.89	108	Theme 4: Regulatory concerns (8/15 participants)
Correlation Strength	-	r = 0.58 *	r = −0.64 *	384	Strong convergence across methods

Note: * p < 0.05. Qualitative validation shows the percentage of interview participants supporting each pattern.

, technological optimism framing dominates Indonesian AI coverage (47.2%). It correlates significantly with higher public confidence in AI systems (r = 0.58, p < 0.01) but paradoxically reduces awareness of AI limitations and risks.

Cross-method validation demonstrates this pattern across multiple data sources. Content analysis reveals government officials (31.2%) and technology industry sources (24.7%) predominantly drive optimistic framing, while survey respondents exposed to such coverage demonstrate elevated confidence scores but lower risk awareness. Figure 1 visualizes these convergent relationships through an integrated model showing how media patterns influence public understanding.

Qualitative analysis through Theme 2 (Editorial Preferences and Institutional Pressures) contextualizes this finding: “We tend to focus on the positive potential of AI because our readers need to understand how technology can help Indonesia’s development goals” (Senior editor, Kompas, 15 September 2024).

The geographic dimension strengthens this convergent pattern. Despite 45% exposure to AI media coverage, rural populations show 67% lower AI risk awareness than their urban counterparts. This finding aligns with Theme 3 (Educational Gaps), where 12/15 interview participants identified “urban-rural disparities in AI understanding” as a critical concern.

Technical accuracy issues compound these effects. Content analysis identifies factual errors in 41.6% of AI articles, while survey data shows only 32.3% of respondents can adequately evaluate AI-generated content. Theme 1 (Professional Challenges) explains this convergence: “We struggle with AI coverage because the technology changes so fast, and we don’t always have access to technical experts” (Technology journalist, Detik.com, 3 October 2024).

4.2. Digital Literacy-Information Resilience Nexus

Multi-source analysis reveals complex interactions between digital literacy capabilities and information resilience that vary significantly across demographic segments.

Table 5. Digital literacy scores by demographic group.

Demographic Segment	Digital Literacy Score	AI Understanding (%)	Information Resilience	Qualitative Insights
Age Groups
18–25 (Urban)	3.42 ± 0.87	45.2%	3.21	Theme 3: Technical skills high, critical evaluation moderate
26–35 (Mixed)	3.18 ± 0.92	38.7%	2.94	Theme 1: Professional pressures affect information processing
36–45 (Mixed)	2.54 ± 1.05	28.3%	2.45	Theme 3: Theory-practice gap most pronounced
46+ (Rural-heavy)	1.89 ± 0.98	15.1%	1.87	Theme 4: Traditional verification skills underutilized
Education
University	3.65 ± 0.76	52.7%	3.34	Theme 3: Formal knowledge does not guarantee AI literacy
High School	2.43 ± 0.89	25.8%	2.18	Theme 2: Media influence is strongest in this segment
Elementary	1.78 ± 0.94	12.4%	1.65	Theme 4: Social networks provide alternative verification
Location
Urban	3.21 ± 0.94	41.8%	2.89	Theme 2: Higher exposure, mixed comprehension
Rural	2.18 ± 1.12	18.3%	2.31	Theme 3: Geographic education gaps are critical
Gap Analysis	1.03 points	23.5%	0.58 points	Consistent across all themes

Note: ± indicates standard deviation. Qualitative insights were synthesized from thematic analysis.

presents an integrated analysis showing how overall digital literacy scores (mean = 2.76) mask substantial variations directly correlating with information vulnerability patterns.

Convergent evidence from survey instruments and qualitative analysis demonstrates systematic gaps between theoretical knowledge and practical AI evaluation skills. While 41.9% of respondents hold a university education, only 32.3% demonstrate adequate understanding of AI decision-making processes.

Figure 2 illustrates the complex multi-dimensional interactions between digital literacy capabilities and information resilience that vary significantly across demographic segments.

Theme 3 (Educational Gaps and Public Understanding), mentioned by 14/15 interview participants, contextualizes this finding: “We see students with strong theoretical knowledge but limited practical skills in evaluating AI-generated content. The gap between formal education and practical needs is significant” (Digital literacy educator, Hasanuddin University, 28 September 2024).

Age-related patterns show strong convergence across data sources. Survey data reveals declining digital literacy with age (18–25: 3.42, 46+: 1.89), while qualitative analysis identifies generational differences in AI comprehension among 9/15 participants.

Geographic disparities demonstrate the most substantial convergent effects. Urban-rural digital literacy gaps (3.21 vs. 2.18 mean scores) amplify information vulnerability, particularly for AI-generated content assessment. This quantitative finding aligns with interview data where policy experts noted: “AI-related education and exposure are concentrated in urban areas, creating new forms of information inequality” (Ministry official, 12 October 2024).

4.3. Media-Literacy-Resilience Interaction Effects

Integrated analysis reveals how media coverage patterns, digital literacy levels, and information resilience interact to create specific vulnerability profiles among Indonesian internet users.

Table 6. Three-way interaction analysis: media exposure × digital literacy × information resilience.

Interaction Profile	Media Exposure Pattern	Digital Literacy Level	Information Resilience	Vulnerability Index	Qualitative Validation
High-Risk Profile	Optimistic-Heavy (>70%)	Low Critical Skills (<2.5)	Low Resilience (<2.0)	4.35 (Very High)	Theme 2 + Theme 3: Rural, older, limited education
Moderate-Risk Profile	Mixed Exposure (40–70%)	Moderate Skills (2.5–3.5)	Moderate Resilience (2.0–3.0)	2.78 (Moderate)	Theme 1 + Theme 4: Urban educated, time-pressured
Lower-Risk Profile	Balanced/Skeptical (>30% cautionary)	High Critical Skills (>3.5)	High Resilience (>3.0)	1.92 (Lower)	All themes: Urban, young, higher education
Protective Profile	Limited Exposure (<30%)	Traditional Verification	Community-Based Resilience	2.15 (Moderate)	Theme 4: Rural communities with strong social networks

Note: Vulnerability Index calculated from a weighted combination of overconfidence, risk awareness gaps, and susceptibility to AI-generated misinformation.

presents three-way interaction analysis showing how these factors combine to create distinct risk profiles.

The most significant interaction occurs between media framing exposure and individual critical evaluation capabilities. Respondents with high exposure to optimistic AI framing but low essential skills of evaluation show the highest overconfidence in AI systems (confidence score 4.12), combined with the lowest risk awareness (1.87).

Source utilization patterns in media coverage correlate with public information evaluation behaviors. Survey respondents who primarily consume government and industry-framed AI coverage show a 43% lower likelihood of seeking alternative perspectives. Theme 2 (Editorial Preferences) explains this pattern through institutional pressure: “Commercial pressure to avoid negative technology coverage” affects 6/15 media professional participants.

Technical journalism errors create cascading effects on public understanding. Content analysis error rates (41.6%) correlate significantly with public misunderstanding patterns (r = 0.64), particularly affecting populations with lower educational attainment. Theme 1 (Professional Challenges) identifies systematic causes: technical complexity barriers (13/15 participants), limited expert access (9/15 participants), and time pressure versus accuracy demands (8/15 participants).

4.4. Cross-Method Validation Matrix

Systematic comparison across quantitative and qualitative data sources demonstrates strong convergence for key findings while revealing nuanced variations.

Table 7. Cross-method validation results with statistical significance.

Primary Finding	Quantitative Evidence	Qualitative Support	Statistical Significance	Convergence Strength	Policy Implications
Media optimism bias creates public overconfidence	47.2% optimistic framing; 3.84 confidence score vs. 2.31 risk awareness; r = 0.58 correlation	Theme 2: Editorial preferences (11/15); “focus on positive potential” quotes from 3 major outlets	p < 0.01; 95% CI [0.42, 0.71]	Strong	Journalist training in balanced reporting
Technical journalism gaps affect public understanding	41.6% article error rate; 32.3% adequate AI comprehension; r = 0.64 correlation	Theme 1: Professional challenges cited by 13/15 participants; specific technical errors documented	p < 0.001; 95% CI [0.51, 0.75]	Very Strong	Technical journalism education, expert access programs
Education-practice gap undermines AI literacy	Digital literacy 2.76 mean; 24.7% AI bias recognition; r = 0.52 correlation	Theme 3: Educational gaps identified by 14/15 participants; “theory-practice gap” consistent	p < 0.05; 95% CI [0.33, 0.68]	Moderate	Curriculum reform, practical AI literacy training
Geographic disparities amplify information vulnerability	Urban 3.21 vs. rural 2.18 literacy scores; 1.52 point AI understanding gap; r = 0.67 correlation	Urban-rural disparities mentioned by 12/15 participants; infrastructure and access issues	p < 0.001; 95% CI [0.54, 0.77]	Strong	Rural digital inclusion programs, targeted interventions
Policy frameworks lag technological adoption	73% need updated regulation; 2.68 information resilience mean; r = 0.45 correlation	Theme 4: Regulatory concerns from 8/15 participants; “regulatory lag” noted by policy experts	p < 0.05; 95% CI [0.28, 0.61]	Moderate	Proactive AI governance, regulatory capacity building

Note: CI = Confidence Interval. Convergence strength based on effect size, significance, and qualitative validation consistency.

provides comprehensive validation results showing statistical significance and convergence strength for each significant finding.

Divergent findings provide additional insights. While quantitative data shows older adults have the lowest digital literacy (46+: 1.89), qualitative analysis reveals they demonstrate stronger source verification behaviors in traditional media contexts. This suggests age-related variations in digital literacy may mask transferable critical thinking skills.

Geographic convergence shows the strongest validation. Quantitative urban-rural gaps (1.52 points for AI understanding) align precisely with qualitative observations about concentrated urban exposure and education. However, qualitative data reveal that rural communities possess strong social verification networks that could be enhanced through targeted interventions.

4.5. Integrated Implications for Information Resilience

Multi-source analysis reveals that Indonesian society faces systemic challenges in developing AI-era information resilience and identifies specific leverage points for intervention.

Table 8. Integrated Intervention Priority Matrix.

Intervention Area	Quantitative Justification	Qualitative Support	Expected Impact	Implementation Complexity	Priority Ranking
Journalism Training	41.6% error rate; r = 0.64 public impact	Theme 1: 13/15 cite technical challenges	High	Moderate	1st Priority
Rural Digital Inclusion	1.52 point literacy gap; 67% lower risk awareness	Theme 3: 12/15 identify geographic disparities	Very High	High	2nd Priority
AI-Specific Education	32.3% adequate comprehension; 24.7% bias recognition	Theme 3: 14/15 note education-practice gap	High	Moderate	3rd Priority
Media Diversity Policy	31.2% government sources; 47.2% optimistic framing	Theme 2: 11/15 describe editorial pressures	Moderate	Low	4th Priority
Regulatory Framework	73% need updated regulation; 2.68 resilience score	Theme 4: 8/15 highlight policy lag	Moderate	High	5th Priority

Note: Priority ranking based on expected impact, implementation feasibility, and convergence strength.

presents an integrated intervention priority matrix based on convergent evidence across methodological approaches.

The combination of optimistic media framing (47.2%), moderate digital literacy (2.76 mean), and concerning AI-specific gaps (32.3% adequate understanding) creates conditions where technological adoption outpaces critical evaluation capabilities. Theme 4 (Policy and Regulatory Concerns) contextualizes quantitative vulnerability scores (2.68 mean information resilience): “We see the public becoming more trusting of AI-powered fact-checking and content moderation without understanding their limitations” (Policy expert, 12 October 2024).

Educational system gaps emerge as the most consistent finding across all data sources. Formal education correlates strongly with digital literacy (r = 0.73), but qualitative analysis reveals systematic theory-practice disconnects. University graduates show technical competency but struggle with AI-specific evaluation tasks, while those with practical technology experience demonstrate better adaptive capabilities regardless of formal education level.

The interaction between media coverage patterns and public understanding suggests targeted interventions could have multiplier effects. Improving technical journalism accuracy (addressing 41.6% error rates) and enhanced critical evaluation education (targeting 32.3% adequate comprehension) could significantly strengthen information resilience across demographic segments.

5. Discussion

5.1. Validated Media-Literacy-Resilience Framework

The triangulated findings validate a comprehensive model where media framing patterns, digital literacy capabilities, and information resilience interact systematically to create specific vulnerability profiles. As demonstrated in Table 7, cross-method validation indicates that these relationships operate consistently across demographic segments while revealing significant contextual variations that will be further illustrated in the comprehensive validation matrix.

Systematic comparison across quantitative and qualitative data sources demonstrates strong convergence for key findings while revealing nuanced variations. Figure 3 provides comprehensive validation results showing statistical significance and convergence strength for each significant finding.

The strongest validated relationship emerges between media framing bias and public overconfidence in AI systems (r = 0.58, p < 0.01), supported by qualitative evidence from 11/15 interview participants who identified institutional pressures driving optimistic coverage. This finding extends media framing theory (Entman, 1993) into the AI domain, showing how journalistic choices about technological coverage create measurable effects on public risk perception.

Convergent evidence across quantitative and qualitative data sources reveals that traditional models of digital literacy prove insufficient for AI-mediated information environments. The education-practice gap identified in Table 5 shows how formal education correlates strongly with general digital literacy (r = 0.73), but Theme 3 findings contextualize this statistical relationship, revealing disconnects between theoretical knowledge and practical AI evaluation skills.

The geographic dimension provides the most robust convergent validation, as shown in Figure 2. Urban-rural disparities appear consistently across content analysis (concentrated coverage in urban-focused outlets), survey data (1.52 point literacy gap), and interview insights (12/15 participants identifying geographic inequalities). This convergence suggests that traditional digital divide frameworks remain relevant but require adaptation for AI-specific competencies.

Information resilience is a multidimensional construct that cannot be predicted solely from digital literacy or media exposure variables. The stronger correlation between critical evaluation skills and information resilience (r = 0.71) compared to technical skills (r = 0.42), as validated in Table 7, aligns with qualitative findings emphasizing analytical capabilities over technical competencies.

5.2. Integrated Policy Implications from Multi-Source Analysis

Convergent evidence across quantitative and qualitative data sources points to specific, actionable intervention strategies that address systemic rather than isolated challenges. Table 8 demonstrates how the integration of findings reveals leverage points where targeted interventions could create cascading improvements across multiple dimensions of information resilience.

Journalism education emerges as the highest-impact intervention opportunity based on convergent analysis. Technical accuracy issues affecting 41.6% of AI coverage correlate directly with public misunderstanding patterns (r = 0.64), while Theme 1 (Professional Challenges) identifies specific, addressable barriers: limited technical expertise (11/15 participants), restricted expert access (9/15 participants), and time pressure conflicts (8/15 participants).

Rural digital inclusion represents the most urgent equity concern, as illustrated in Figure 2 and validated in Table 7. Triangulation reveals how geographic disparities compound across multiple dimensions: media coverage concentration, educational access limitations, and infrastructure constraints. However, qualitative analysis also identifies protective factors in rural communities, particularly strong social verification networks that could be enhanced rather than replaced.

Educational curriculum reform shows strong convergent justification but requires nuanced implementation. The theory-practice gap identified by 14/15 interview participants aligns with survey findings shown in Table 5, revealing a disconnect between formal education levels and AI-specific competencies. Successful interventions must address both institutional educational structures and informal learning mechanisms.

Regulatory frameworks require proactive development rather than reactive responses. Theme 4 (Policy and Regulatory Concerns) converges with quantitative findings in Table 7, showing moderate information resilience (2.68 mean) and high perceived need for updated regulation (73% of respondents). However, regulatory approaches must account for the complex interaction effects revealed through mixed-methods analysis.

Integration Across Intervention Areas

The convergent analysis reveals that isolated interventions are likely to prove insufficient given the systemic nature of identified challenges, as demonstrated in the interaction effects shown in Table 6. Media training without addressing educational gaps may improve coverage quality without enhancing public critical evaluation capabilities. Similarly, educational interventions without corresponding improvements in media coverage quality may create more informed audiences for persistently problematic information sources.

Successful intervention strategies should address multiple system components simultaneously. Theme integration suggests that journalism training programs should include collaboration with educational institutions, while educational reforms should incorporate media literacy components specifically addressing AI technologies. Rural inclusion initiatives should leverage existing social networks while building technological infrastructure and capabilities.

5.3. Theoretical Contributions of Convergent Mixed-Methods Approach

This research demonstrates how a convergent mixed-methods design enables a deeper understanding of complex socio-technical phenomena than single-method approaches. The triangulation of content analysis, survey research, and expert interviews reveals interaction effects and contextual factors that would remain invisible through isolated quantitative or qualitative investigation, as evidenced in Figure 3’s convergence validation.

Media framing theory gains significant extension through this integrated approach. While traditional framing analysis focuses on content characteristics and audience reception separately, the convergent design shown in Figure 1 demonstrates how framing effects operate through interaction with individual digital literacy capabilities and demographic characteristics. The finding that framing effects vary significantly across literacy levels (r = 0.43 in urban vs. r = 0.71 in rural contexts) provides new theoretical insights.

Digital divided literature benefits from the multidimensional analysis presented in Table 5 and Figure 2, revealing how traditional access and usage metrics mask more fundamental competency gaps. The convergent evidence shows that AI-era digital divides operate through complex interactions between technological access, educational preparation, media exposure patterns, and critical evaluation capabilities.

Information resilience emerges as a distinct theoretical construct requiring integration across media studies, education research, and technology adoption frameworks. The validated relationships between critical evaluation skills, information resilience, and vulnerability profiles shown in Table 6 and Table 7 provide an empirical foundation for theoretical development in this emerging area.

Methodological contributions include demonstration of how qualitative findings can contextualize and explain quantitative relationships, while quantitative analysis can validate and generalize qualitative insights. The cross-method validation matrix in Table 7 provides a replicable framework for assessing convergence strength and identifying areas where different methodological approaches provide complementary rather than redundant insights.

Limitations and Methodological Considerations

Several limitations should be acknowledged despite the strengths of the convergent approach demonstrated in Figure 3. The cross-sectional design limits causal inferences about relationships between media exposure and digital literacy development. ==While providing detailed local insights, the geographic focus on South Sulawesi may limit generalizability to other Indonesian regions with different cultural and economic characteristics.

The sample size of 384 respondents, while statistically adequate as shown in Table 5, represents a relatively small proportion of the Indonesian internet-using population. Future research with larger, nationally representative samples would strengthen the generalizability of these findings. However, qualitative validation through expert interviews provides additional confidence in the broader applicability of identified patterns, as demonstrated in the convergence validation presented in Table 7.

6. Conclusions

This convergent mixed-methods analysis provides robust evidence for systematic relationships between AI media coverage, digital literacy, and information resilience in Indonesia. Triangulation across content analysis, survey research, and expert interviews validates key findings while revealing nuanced interactions that single-method approaches might miss, as demonstrated throughout Table 4, Table 5, Table 6, Table 7 and Table 8 and Figure 1, Figure 2 and Figure 3.

The research reveals significant challenges and opportunities in the relationship between AI media coverage and digital literacy in Indonesia, aligning with global patterns identified in similar studies (Newman, 2024; Helberger, 2019). The predominance of optimistic framing in media coverage (47.2%), combined with moderate digital literacy levels (mean = 2.76) and concerning gaps in AI-specific understanding (32.3% adequate comprehension), creates conditions where Indonesian society may be inadequately prepared for AI-mediated information environments, as illustrated in Figure 1.

The findings demonstrate that media narratives about AI technologies significantly influence public understanding and attitudes, but current coverage patterns may not optimally serve public education and democratic deliberation needs. The technical inaccuracies found in 41.6% of analyzed articles and the vulnerability profiles identified in Table 6 suggest that improvements in technology journalism practices and professional development initiatives are urgently needed.

Digital literacy levels among Indonesian internet users show concerning variations across demographic segments, with vulnerabilities among rural populations (mean = 2.18), older adults, and individuals with limited formal education, as detailed in Table 5. The gaps in AI-specific digital literacy are especially problematic given the increasing prevalence of AI systems in information production and distribution. Cross-method validation confirms that these quantitative patterns reflect deeper systemic challenges in educational preparation and professional capacity, as validated in Table 7.

Information resilience levels reflect these digital literacy patterns while demonstrating the complex relationships between competency areas and actual information evaluation behaviors, as shown in Figure 2. The stronger correlation between critical evaluation skills and information resilience (r = 0.71) compared to technical skills (r = 0.42) suggests that educational interventions should prioritize analytical capabilities over purely technical training.

The policy implications of these findings are substantial, pointing to urgent needs for educational reform, journalist training, and regulatory development that can help Indonesian society navigate AI-mediated information environments more effectively. Table 8 presents a prioritized intervention framework where the success of these interventions will depend on recognition of demographic variations and the need for tailored approaches that build on existing strengths while addressing specific vulnerabilities.

Convergent evidence demonstrates that isolated interventions are likely insufficient given the systemic nature of identified challenges. As shown in the interaction analysis presented in Table 6, media training without addressing educational gaps may improve coverage quality without enhancing public critical evaluation capabilities. Educational interventions without corresponding improvements in media coverage quality may create more informed audiences for persistently problematic information sources.

As Indonesia continues its digital transformation journey, the relationship between media coverage, digital literacy, and information resilience will remain crucial for maintaining democratic discourse and informed decision-making processes. The challenges identified in this research are significant but not insurmountable, provided that appropriate attention and resources are directed toward addressing them through coordinated efforts across media, educational, and policy sectors.

The implications of this research extend beyond Indonesia to other developing digital economies facing similar challenges in balancing rapid technological adoption with information integrity and democratic participation. The patterns identified in Indonesian media coverage and public digital literacy reflect broader global trends while maintaining important local characteristics that inform context-specific intervention strategies.

The convergent mixed-methods approach validates the value of integrating multiple data sources and analytical perspectives to understand complex socio-technical phenomena, as demonstrated through Figure 3’s validation framework. This methodological framework provides a replicable model for investigating similar challenges in other contexts while demonstrating how quantitative and qualitative insights can strengthen each other through systematic triangulation.

Future research should focus on longitudinal analysis to understand how these relationships evolve as AI technologies and public understanding develop. Additionally, comparative analysis across multiple Indonesian regions and other Southeast Asian countries would provide valuable insights into how cultural, economic, and political factors influence the dynamics identified in this study.

The fundamental challenge for Indonesian society and others facing similar transformations lies in developing information resilience capabilities that match the pace of technological change while preserving democratic values and social cohesion. Meeting this challenge requires sustained commitment to evidence-based policy development, educational innovation, and media system improvements that serve the public interest in an increasingly complex information landscape.